USF Libraries
USF Digital Collections

Developing a school social work model for predicting academic risk

MISSING IMAGE

Material Information

Title:
Developing a school social work model for predicting academic risk school factors and academic achievement
Physical Description:
Book
Language:
English
Creator:
Lucio, Robert
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Educational outcomes
School related factors
Ecological perspective
Cumulative risk
Additive risk
Dissertations, Academic -- Social Work -- Doctoral -- USF   ( lcsh )
Genre:
non-fiction   ( marcgt )

Notes

Abstract:
ABSTRACT: The impact of school factors on academic achievement has become an important focus for school social work and revealed the need for a comprehensive school social work model that allows for the identification of critical areas to apply social work services. This study was designed to develop and test a more comprehensive school social work model. Specifically, the relationship between cumulative grade point average (GPA) and the cumulative risk index (CRI) and an additive risk index (ARI) were tested and a comparison of the two models was presented. Over 20,000 abstracts were reviewed in order to create a list of factors which have been shown in previous research to impact academic achievement. These factors were divided into the broad domains of personal factors, family factors, peer factors, school factors, and neighborhood or community factors. Factors that were placed under the school domain were tested and those factors which met all three criteria were included in the overall model. Consistent with previous research, both the CRI and ARI were shown to be related to cumulative GPA. As the number of risk factors increased, GPA decreased. After a discussion of the results, a case was made for the use of an additive risk index approach fitting more with the current state of social work. In addition, selecting cutoff points for determining risk and non-risk students was accomplished using an ROC analysis. Finally, implications for school social work practice on the macro-, meso-, and micro- levels were discussed.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2008.
Bibliography:
Includes bibliographical references.
System Details:
Mode of access: World Wide Web.
System Details:
System requirements: World Wide Web browser and PDF reader.
Statement of Responsibility:
by Robert Lucio.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 171 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002004998
oclc - 355371546
usfldc doi - E14-SFE0002699
usfldc handle - e14.2699
System ID:
SFS0027016:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 002004998
003 fts
005 20090528124458.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 090528s2008 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0002699
040
FHM
c FHM
035
(OCoLC)355371546
049
FHMM
090
(Online)
1 100
Lucio, Robert.
0 245
Developing a school social work model for predicting academic risk :
b school factors and academic achievement
h [electronic resource] /
by Robert Lucio.
260
[Tampa, Fla] :
University of South Florida,
2008.
500
Title from PDF of title page.
Document formatted into pages; contains 171 pages.
Includes vita.
502
Dissertation (Ph.D.)--University of South Florida, 2008.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
3 520
ABSTRACT: The impact of school factors on academic achievement has become an important focus for school social work and revealed the need for a comprehensive school social work model that allows for the identification of critical areas to apply social work services. This study was designed to develop and test a more comprehensive school social work model. Specifically, the relationship between cumulative grade point average (GPA) and the cumulative risk index (CRI) and an additive risk index (ARI) were tested and a comparison of the two models was presented. Over 20,000 abstracts were reviewed in order to create a list of factors which have been shown in previous research to impact academic achievement. These factors were divided into the broad domains of personal factors, family factors, peer factors, school factors, and neighborhood or community factors. Factors that were placed under the school domain were tested and those factors which met all three criteria were included in the overall model. Consistent with previous research, both the CRI and ARI were shown to be related to cumulative GPA. As the number of risk factors increased, GPA decreased. After a discussion of the results, a case was made for the use of an additive risk index approach fitting more with the current state of social work. In addition, selecting cutoff points for determining risk and non-risk students was accomplished using an ROC analysis. Finally, implications for school social work practice on the macro-, meso-, and micro- levels were discussed.
538
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
590
Co-advisor: William Rowe, Ph.D.
Co-advisor: Lisa Rapp-Paglicci, Ph.D.
653
Educational outcomes
School related factors
Ecological perspective
Cumulative risk
Additive risk
690
Dissertations, Academic
z USF
x Social Work
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.2699



PAGE 1

Developing a School Social Work Model for Predicting Academic Risk : School Factors and Academic Achievement by Robert Lucio A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy School of Social Work College of Behavioral and Community Sciences University of South Florida Co Major Professor: William Rowe, D.S.W Co Major Professor: Lisa Rapp Paglicci, Ph.D. Chris Stewart, Ph.D. Shannon Suldo, Ph.D. Date of Approval: October 21, 2008 Keywords: educational outcomes, school related factors ecological perspective, cumulative risk, additive risk ROC Curve Analysis risk factors, protective factors Copyright 2008 Robe rt Lucio

PAGE 2

Acknowledgement I would like to express my sincerest gratitude to a host of people who inspired me to achieve my goals, grow, and most of all, to complete my educational journey. I have been honored to have been given the opportunity to work with so many great people dur ing this process. I would like to personally thank Dr. William Rowe, Dr. Lisa Rapp Paglicci, Dr. Chris Stewart, and Dr. Shannon Suldo for helping lead me through this process and encouraging me to do the best work I could. Your guidance has been invaluable and I have learned so much through this process, not only about social work but about myself. I would also like to offer my appreciation for all of the amazing professors I have had duri ng my doctoral education; Dr. Mary Armstrong, Dr. Roger Boothroyd, Dr. Robin Ersing, Dr. Jerry Miller, Dr. Greg Paveza, Dr. Michael Rank, and Dr. Cleo Roberts. To my classmates in the first graduating social work Ph.D. class at the University of South F lorida. I was blessed to have such a diverse and talented group of people to be around for 4 years of my life. I hope the opportunity arises that we will be able to collaborate on future work. I would also like to extend a special thanks to Don and Brig ette for their unbelievable friendship support, and help through out this process. Finally to the PCS s ocial w orkers and the administration and staff at my research schools. Your dedication as social workers, teachers and professionals is truly amazing and your help was invaluable I would also like to extend my thanks to those students who participated in my study, sharing their thoughts and f eelings with me for my research as nothing would be possible without their help.

PAGE 3

Dedication To my wonder fully patient wife, Maridelys. Your constant support and belief in me has helped maintain my strength and ability to persevere. Not to mention dealing forward to many dissertation free days. You have been my guiding light and I could never express how much your love and support has meant to me. To my parents, Mammy and Drrr. From an early age you have always stressed the importance of education and encouraged me to strive to do the best I can. No matter what I have chosen to do, you have always been behind me 100 percent of the time. Thank you for everything.

PAGE 4

i Table of Contents Table of Contents ................................ ................................ ................................ ................. i List of Tables ................................ ................................ ................................ ...................... v List of Figures ................................ ................................ ................................ .................... vi Abstract ................................ ................................ ................................ ............................. vii Chapter One: Introduction ................................ ................................ ................................ .. 1 Purpose of Study ................................ ................................ ................................ ..... 7 Dissertation Questions ................................ ................................ ............................ 8 Chapter Two: Literature Review ................................ ................................ ........................ 9 The Ecological Perspective ................................ ................................ ..................... 9 Ecological Model and Risk ................................ ................................ ....... 11 Ecological Model and School Social Work ................................ .............. 12 The Need for a More Comprehensive Model ................................ ........... 13 Cumulative Risk ................................ ................................ ................................ .... 15 Empirical Support for the Cumulative Risk Model ................................ .. 17 Additive Model ................................ ................................ ................................ ..... 19 Empirical Support for the Additive Model ................................ ............... 20 Risk and Protective Factors ................................ ................................ .................. 21 Risk Factors ................................ ................................ .............................. 23 Protective Factors ................................ ................................ ...................... 23 Model Development and Testing ................................ ................................ .......... 25 Outcome Variable: Grade Point Average (GPA) ................................ .................. 25 Domains ................................ ................................ ................................ ................ 26 School Related Factors ................................ ................................ ......................... 27 Concept Mapping ................................ ................................ ................................ .. 28 Concept Mapping Activity ................................ ................................ .................... 30 School Factors ................................ ................................ ................................ ....... 32 Academic Engagement ................................ ................................ ............. 33 Academic Expectations (Student) ................................ ............................. 34 Academic Self Efficacy ................................ ................................ ............ 35 Attendance ................................ ................................ ................................ 35 Educational Support ................................ ................................ .................. 36 Grade Retention ................................ ................................ ........................ 37 Homework ................................ ................................ ................................ 38 Music Instruction ................................ ................................ ...................... 38 School Behaviors ................................ ................................ ...................... 39

PAGE 5

ii School Belonging ................................ ................................ ...................... 40 School Mobility ................................ ................................ ........................ 41 School Relevance ................................ ................................ ...................... 41 School Safety ................................ ................................ ............................ 42 Supportive School Environment ................................ ............................... 43 Teacher Support/Relationships ................................ ................................ 43 Limitations of Previous Research ................................ ................................ ......... 45 Number of variables ................................ ................................ .................. 45 Self Report Grades ................................ ................................ .................... 46 Analysis Tools ................................ ................................ .......................... 47 Sample Population ................................ ................................ .................... 47 Sample Size ................................ ................................ ............................... 48 Summary ................................ ................................ ................................ ............... 48 Chapter Three: Methodology ................................ ................................ ........................ 51 Participants ................................ ................................ ................................ ............ 51 Response Rates ................................ ................................ ................................ ..... 51 Research Design ................................ ................................ ................................ .... 52 Controlled Factors ................................ ................................ ................................ 52 Instrument ................................ ................................ ................................ ............. 53 Dependent Variable ................................ ................................ .............................. 53 Academic Achievement ................................ ................................ ............ 53 Controlled Factors ................................ ................................ ................................ 53 SES 53 Race 54 Gender 54 School related factors ................................ ................................ ............................ 54 Academic Engagement ................................ ................................ ............ 54 Academic Expectations ................................ ................................ ............ 54 Academic Self Efficacy ................................ ................................ ........... 55 Attendance ................................ ................................ ................................ 5 5 Educational Support ................................ ................................ .................. 56 Grade Retention ................................ ................................ ........................ 57 Homework ................................ ................................ ................................ 57 Music Instruction ................................ ................................ ...................... 57 School Behavior ................................ ................................ ....................... 57 School Belonging ................................ ................................ ..................... 58 School Mobility ................................ ................................ ....................... 58 School Relevance ................................ ................................ ..................... 58 School Safety ................................ ................................ ........................... 59 Teacher Support/Relationship ................................ ................................ .. 59 Instrument Pilot Testing ................................ ................................ ........................ 60 Overview of Risk and Protective Scores ................................ .............................. 61 Risk/Protective Factor Scores ................................ ................................ ... 61 Cumulative Risk Model ................................ ................................ ........................ 61 Additive Risk Model ................................ ................................ ............................. 62

PAGE 6

iii Data Collection Procedure ................................ ................................ .................... 62 Chapter Four: Statistical Analysis ................................ ................................ .................... 64 Analytic Approach ................................ ................................ ................................ 64 Descriptive Analysis ................................ ................................ ............................. 65 Relationships among Factors and Cumulative GPA ................................ ............. 66 Question 1 Individual Risk and Protective Factors ................................ ............ 68 Data Ag gregation ................................ ................................ ...................... 70 Socioeconomic Status (SES) ................................ ......................... 70 Academic Expectations ................................ ................................ 70 Academic Self Efficacy ................................ ................................ 71 Attendance Rate ................................ ................................ ............ 72 Educational Support ................................ ................................ ...... 72 Homework ................................ ................................ ..................... 73 Grade Retention ................................ ................................ ............ 73 Music Playing ................................ ................................ ............... 74 School Behaviors ................................ ................................ .......... 74 School Belonging ................................ ................................ .......... 75 School Mobility ................................ ................................ ............ 75 Controlled Factors ................................ ................................ ......... 76 Unique Contribution of Factors ................................ ................................ 76 Cumulative GPA ................................ ................................ ....................... 77 Computation of the Cumulative Risk Index ................................ ............. 77 Computation of the Additive Risk Index ................................ .................. 78 Descriptive Analysis of the CRI and ARI Factors by Ri sk Group ........... 79 Question 2 Cumulative Risk Model ................................ ................................ ... 79 Question 3 Additive Risk Model ................................ ................................ ........ 81 Question 4 Cumulative Risk Model versus Additive Model ............................. 83 Cross Validation ................................ ................................ ........................ 83 Model Comparison ................................ ................................ .................... 84 Question 5 Differentiation between At risk and non At risk Students .............. 84 C hapter Five: Discussion ................................ ................................ .............................. 86 Findings ................................ ................................ ................................ ................. 86 Model Comparison ................................ ................................ .................... 87 Final Model ................................ ................................ ............................... 89 Indices versus Direct Factors ................................ ................................ .... 90 Factors and Odds of Passing ................................ ................................ ..... 90 Cutoff Points ................................ ................................ ............................. 90 Strengths and Limitations ................................ ................................ ..................... 91 Future Research Directions ................................ ................................ ................... 93 Implications for Social Work Practice ................................ ................................ .. 94 Macrosystem ................................ ................................ ............................. 94 Mesosystem ................................ ................................ ............................... 96 Microsystem ................................ ................................ .............................. 97 Conclusion ................................ ................................ ................................ ............ 98

PAGE 7

iv References ................................ ................................ ................................ ....................... 101 Appendices ................................ ................................ ................................ ...................... 126 Appendix A: Tables and Figures ................................ ................................ ........ 127 Appendix B: Concept Mapping For m ................................ ................................ 161 Appendix C: Survey Instrument ................................ ................................ ......... 162 Appendix D: Consent and Assen t Forms ................................ ............................ 165 Appendix E: Institutional Approval Letter ................................ ......................... 171 About the Author ................................ ................................ ................................ .. End Page

PAGE 8

v List of Tables Table 1. Articles Examined for Literature Review ................................ ......................... 128 Table 2. Factors Related to Academic Achievement ................................ ...................... 129 Table 3. Factors Above 75% Initial Agreement ................................ ............................. 130 Table 4. Domain and Individual Factors ................................ ................................ ......... 132 Table 5. Demographic Characteristics of Sample ................................ ........................... 134 Table 6. Descriptive Information for Study Fact ors ................................ ....................... 135 Table 7. Sample versus Population for Demographic and Control Information ............ 136 Table 8. Correlations Among Risk and Promotive Factors ................................ ............ 137 Table 9. Unique Contribution of CRI Factors Regression Model ................................ .. 138 Table 10. Unique Contribution of ARI Factors Regression Model ................................ 139 Table 11. Descriptive Information for Dependent and Independent Factors .................. 140 Table 12. ARI Risk, Non Risk, and Promotive Factors ................................ ................. 141 Table 13. CRI Risk Factors ................................ ................................ ............................. 142 Table 14. Correlation of Regression Factors for the CRI ................................ ............... 146 Table 15. Standard Multiple Regression Results for the CRI ................................ ......... 147 Table 16. Correlation of Regression F actors for ARI Model ................................ ......... 152 Table 17. Standard Multiple Regression Results for ARI Model ................................ ... 153 Table 18. Logistical Regression Results for CRI and ARI models. ............................... 155 Table 19. Area Under the Curve and Accuracy Indices for CRI and ARI ..................... 156

PAGE 9

vi List of Figures Figure 1. Broad Ecological Model ................................ ................................ .................. 127 Figure 2. Multi dimensional Scaling Plot ................................ ................................ ....... 131 Figure 3. Initial Ecological Model of School Related Factors ................................ ........ 133 Figure 4. Scatter plot of Residuals for CRI Model ................................ ......................... 143 Fi gure 5. Linearity of Residuals for CRI Model ................................ ............................. 144 Figure 6. Distribution of Residuals For CRI Model ................................ ....................... 145 Figure 7. Cumulative GPA by CRI Scores ................................ ................................ ..... 148 Figure 8. Scatter plot of Residuals for ARI Model ................................ ......................... 149 Figure 9. Linearity of Residuals for ARI Model ................................ ............................. 150 Figure 10. Distribution of Residuals for ARI Model ................................ ...................... 151 Figure 11. Cumulative GPA by ARI scores ................................ ................................ .... 154 Figure 12. ROC Curve for the CRI ................................ ................................ ................. 157 Figure 13. ROC Curve for the ARI ................................ ................................ ................. 158 Figure 14. Final Model of School Factors and Cumulative GPA ................................ ... 159 Figure 15. Final Ecological Model ................................ ................................ ................. 160

PAGE 10

vii Developing a School So cial Work Model for Predicting Academic Risk: School Factors and Academic Achievement Robert Lucio ABSTRACT The impact of school factors on academic achievement has become an important focus for school social work and revealed the need for a comprehensive school social work model that allows for the identification of critical areas to apply social work services. This study was designed to develop and test a more comprehensive school social work model. Specifically, the relationship between cumulative grade point average (GPA) and the cumulative risk index (CRI) and an additive risk index (ARI) were tested and a comparison of the two models was presented. Over 20,000 abstracts were reviewed in order to create a list of factors which have been shown in previous research to impact academic achievement. These factors were divided into the broad domains of personal factors, family factors, peer factors, school factors, and neighborhood or community factors. Factors that were placed under the school domain were tested and those factors which met all three criteria were included in the overall model. Consistent wit h previous research, both the CRI and ARI were shown to be related to cumulative GPA. As the number of risk factors increased, GPA decreased. After a discussion of the results, a case was made for the use of an additive risk index approach fitting more with the current state of social work. In addition, selecting cutoff points for determining risk and non risk students was accomplished using an ROC analysis. Finally, implications for school social work practice on the macro meso and micro levels w ere discussed.

PAGE 11

1 Chapter One Introduction The effect of poor achievement and /or not graduating from school has a deep impact on not only the student s themselves, but the entire nation A high school drop out not only earns $9,245 per year less than a hi gh school graduate, but the estimated increase in cost to public welfare and crime is close to $24 billion (Thorstensen, 2004) In 2002, the unemployment rate of African Americans between 20 24 years old with no high school diploma was 32% compared to 6% for those with a college degree or higher, and overall fewer than 40% of drop outs are employed compared to 60% of high sch ool graduates and over 80% of college graduates (Alliance for Excellent Education, 2003c) But economic proble ms are not the only consequence to high school drop outs. The rate of high risk behaviors such as premature sexual activity, early pregnancy, delinquency, crime, violence, alcohol and drug abuse, and suicide has been found to be significantly higher among dropouts (Woods, 1994) When looking at incarceration, 75% of state prison inmates in the United States are drop outs and dropping out increases the odds of being arrested during a lifetime by over 350% (Harlow, 2003) In fact according to the Alliance for Excellent Education (2003a) a simple 1% increase in high school graduation rates w ould save over $1.4 billion dollars in incarceration costs and a one year increase in education would reduce arrests by over 11%. When looking at health costs, teen girls who score in the bottom 20% in reading and math scores are 5 times more likely to be come pregnant than girls in the top 20% (Alliance for Excellent Education, 2003b) In addition male and female student s with low academic achievement are twice as likely to become parents by their senior year in

PAGE 12

2 high school when compared to high achieving students. This is clearly a social problem ployment, and every other aspect of our society. The U.S. Department of Education reports some surprising statistics when looking at the state of education in the United States (Laird, DeBell, & Chapman, 2006; U.S. Department of Edcuation, 2006) They reported that 10.3% of all students will drop out of school w hile close to a quarter of Hispanic students (23.8%) will drop out. Overall, only 75% of freshman will graduate on time, within 4 years of starting high school. Comparing students nationwide, 27.4% of 8 th graders fall below basic reading proficiency and that number climbs to 30% when examining 8 th grade mathematics scores In addition to academic concerns, students are faced with a host of non acad emic challenges at school which can impact achievement related outcom es. Recent statistics indicate that 53 out of every 1000 students experience some type of theft or victimization at school, and 81% of all schools experienced at least one violent incid ent (Dinkes, Cataldi, & Kena, 2006) Additionally, a full 25% of students reported that drugs were made available to them at school, and 28% were bullied in the previous six months. S tudents (National Research Council, 2004, p. 145) Educational a chievement is a consequence of a variety of factors, including family, community, school, peer s and individual factors. As these factors interact with each other, the resulting academic success or failure is the product of a complex, interconnected relationship (Dimmitt 2003) Numerous studies have identified variables

PAGE 13

3 which appear to be related to academic failure, however, there appears to be little agreement by educators, parents, and/or researchers about which specific factors contribute to the student achieveme nt (Aviles, Anderson, & Davila, 2006) In addition, Jimerson, Egeland, Sroufe, and Carlson (2000) argue d that negative academic outcomes rarely occur without warning, but rather are a process that starts early and impacts later development This suggests that if thi s process can be identified early enough the path to academic failure can be altered By understanding which factors impact achievement school social workers can employ services that can promote achievement and reduce potential risk. Consequently, there is a need for an accurate assessment of factors that students are experiencing which are related to academic success and failure. Particularly needed is a model that can focus on a mult itude (Sameroff, 1985, p. 22) A thorough assessment of factors which n and intervention with students at risk for failure, before they leave school. While the history of school social work over the first 100 years has shown adaptability to the changing social climate, it also reveals a process of specialization. Bartlett (1959) made a distinction between generic social work theory and specialized soci al work theory. Accordingly generic social work was for all social workers and each field of social work also needed their own specialization specific theory. Sc hool social work has its roots in the ecological perspective as a conceptual framework for practice (Garrett, 2007b; Germain, 2006) The National Association of Social Workers (NASW) goes on to define the ecolo

PAGE 14

4 and their environment (200 2) This perspective takes into account the person and their environment as influencing factors in any situation and ecological theory has come to define the profession. This means that school social workers assess and intervene with students, families schools, communities and agencies. When merging the ecological model with school social work practice this model is often presented in terms of a risk, protection and resiliency perspective (Garrett, 2007b) Fraser, Richman and Galinsky (1 999) define risk as a probability that a certain event will occur, given a set of specific conditions. Risk factors are those attributes or variables that increase the likelihood that people with similar characteristics will develop a problem. Risks c an be non specific and generic attributes such as child abuse and poverty, or more s pecifi c, such as unskilled parenting. Countering these forces, are protective factors and resiliency attributes. Protective factors modify the risk and can directly reduce the risk of a disorder or problem (Masten & Obradovic, 2006; Sameroff, 2006) These are seen as distinct constructs, though less is known about them. As with risk factors, protective factors can be generic (regularly attending church) or specific (parental supervision) (Fraser et al., 1999; Rutter, 1987) Finally, resiliency has been used to describe persons who adapt to extraordinary circumstances to excel and achieve p ositive outcomes despite the negative circumstances (Masten & Coatsworth, 1998; Rutter, 2006) Not all children who encounter problems in their lives have academic difficulties, as some children continue to functioning competently academic ally d espite exposure to multiple risks (Kennedy & Bennett, 2006; Rutter, 1979)

PAGE 15

5 The use of a risk, protection and resiliency model has been applied to many arenas including childhood in general (Fraser, Kirby, & Smokowski, 2004) child maltreatment (Thomlison, 2004) alcohol and other drugs (Jenson, 2004) failure to thrive (Kerr, Black, & Krishnakumar, 2000) crime (Miller & MacIntosh, 1999) and delinquency (Williams, Ayers, Van Dorn, & Arthur, 2004) There have also been previous researchers who have related the ecological model to school fai lure (Fraser et al., 1999; Richman, Bowen, & Woolley, 2004) While these researchers have discussed numerous components of the risk, protection and resiliency perspective, all of these factors need to be brought under one cohesive academic outcomes focused model. As ome (Lee, 2007, p. 53) it is necessary to bring together all of the factors specifically related to academic achievement. Th e next section will synthesize the comprehensive collection of previous research on factors connected to academic achievement and present this information as a larger ecological model related exclusively to educ ational outcomes for students. Both cumulative risk and cumulative protective may increase or buffer many kinds of problems. The questions in this study will be examined using both the cumulative risk and additive risk approaches. Cumulative risk is based on the notion that no single risk factor is more impactful than any other (Masten & Coatsworth, 1998; Rutter, 1987) In fact, it is the cumulative risk experience that is most important in shaping outcomes. The more risk factors a student has present, the greater the likelihood of experiencing difficulties. An additive risk approach is similar to a cumulative risk approach, except that both risk and protective factors are considered. In an additive risk

PAGE 16

6 model, it is the combine d impact of both risk factors a nd protective factors that must be considered in relation to outcomes (Luthar, 1991; Sameroff, 1985) This model creates an overall experience of risk and protection which includes risks, assets, and protective processes. More risk factors present will create greater risk, but this risk experience is counterbalanced by protective mechanisms. The ability to identify students who are at most risk for failing is a vital role for school social workers. A thorough view based on the ecological model is essential to accomplish th is, yet there is currently no comprehensive model to utilize. Instead there are an abundance of facto r s which comprise known risk and protective factors in each domain, with little weight or context provided for each As school social workers find themse lves in the fight for limited resources, it is also vital that any school social work model be geared toward academic achievement. School social workers have the ability to influence a reas which will improve student s achievement, behaviors, and other sch ool outcomes but currently have no way of showing this to those who are making the financial decisions School districts and school personnel need to be shown that hiring school social workers can impact student achievement as much as other potential res ources. As schools decide where to spend the limited amount of money they have access to, it is vital for the profession to begin to demonstrate that not only are the For instance, providing comprehensive counseling programs has been shown to improve achievement related outcomes for students (Lapan, Gysbers, & Petroski, 2001; Powell & Arriola, 2003; Sink & Stroh, 2003) which is one example of the impact school social workers can have on student outcomes.

PAGE 17

7 The ability to ide ntify specific risk and promotive factors can help identify places where social workers can intercede. These interventions can take place at a univers al, selective, or indicated level (Bureau of Exceptional Education and Student Services, 2 006; Ezpelata, Granero, de la Osa, & Domenech, 2008; Glover & Albers, 2007; Gordon, 1987) This study will attempt to draw together the current research on academic risk and protective factors and integrate these factors into a school social work model developed to predict academic achievement through school related factors. Purpose of Study The goal of this study will be to identify the factors associated with academic achievement as identified though previous research and draw them together under one comprehensive model. This will be done in the context of the ecological perspective, which allows look ing beyond the in dividual and including environmental factors which interact with t he individual student to create, m aintain reduce, or eliminate academic difficultie s. While each of these factors can be grouped into one of the following domains; personal characteristics, family, peers and friends, school, and neighborhood/community, this dissertation will focus specifically on the school factors domain. In this frame work those factors that represent the school domain will be examined in more detail with an emphasis on the relationship to academic achievement. A final school social work specific model will be presented with an examination of which school related fact ors impact achievement and if a cumulative risk or additive model presents a more accurate prediction of achievement.

PAGE 18

8 Dissertation Questions 1. Which school factors impact academic achievement among high school students? 2. Does the cumulative risk model predict academic achievement among high school students? 3. Does the additive risk model predict academic achievement among high school students? 4 Is the cumulative risk model or additive model a better p redictor of achievement levels for high school students? 5 What is the optimal number of academic domain risk factors for distinguishing between students who are at risk and not at risk?

PAGE 19

9 Chapter Two Literature Review The Ecological Perspective Approaching any problem from an ecological perspective involves viewing interaction s between the person and his/her environment. From this theoretical approach, the environment is comprised of four levels, each defined by the proximity to the individual child (Bronfenbrenner, 1979) These levels are defined by their increasing distance from the individual child and include the microsystem, the mesosystem, the exosystem, the chronosystem, and the macrosystem Individual risk and protective factors occur at each level and in some cases the impact crosses through several levels. For instance, laws providing more pay could cause a mother to work less, which in turn allows her to spend time with her child, finally leading to a better mother child relationship. The National Association of Social Workers (2002) provides a broad nd family and their environment The microsystem is the fir st and most proximal level to the child. This is defined (Richman et al., 2004, p. 146) and includes individual characteristics and individual factors. According to Bronfenbrenner (1979) this would include not only activities and relationships, but als o neighborhoods, family, friends, peers, and schools. After almost 10 years, Bronfenbrenner (1989) revised the definition of the microsystem to include developmentally relevant characteristics of other persons within the environment. This

PAGE 20

10 change was included to account for the interaction that an individual has with other persons in the environment and the influence they may present. Specifically, the revised definition reads: A microsystem is a pattern of activities, roles, and interpersonal relations experienced by developing person in a given face to face setting with particular physical and material features, and containing other persons with distinctive characteristics of temperament, personality and systems of belief (p. 227). The second level away from the child is the mesosystem which comprises the interactions, linkages, a nd processes that take place between two or more settings (Bronfenbrenner, 1989) This second layer refers to the social structures t hat affect, but do not directly include the student. Simply put, this is a system of microsystems. Within this context, this could include a history of school problems or even the relations hip between home and school. The third level of the ecological m odel is the e xosystem which includes linkages and processes that take place between two settings. This includes settings "that do not involve the developing person as an active participant, but in which events occur that affect, or are affected by, what is happening in the setting containing the developing person" (Bronfenbrenner1979, p. 25) Corcoran Franklin and Bennett (2000) interpret t he exosystem as moving to environmental groups that the person does not interact with directly, but still have influences on the person including school administr are responsible for administering educational p rograms which may affect the student. The macrosystem involves the broad distal variables that are farthest away from the child and is a system in which the child does not directly interact (Bronfenbrenner, 1979; Fraser et al., 2004) This level of interaction involves the broa d political, economic, cultural values, beliefs, ideologies and institutional levels of society.

PAGE 21

11 Relationships to the child at this level could involve laws that are passed which open or limit opportunities and that affect the other systemic levels, such as families, schools or communities. Along with the microsystem, the definition of the macrosystem was also revised by Bronfenbrenner (1989) after noting the need to include a dynamic component. This is has been clari fied as: The macrosystem consists of the overarching pattern of micro meso and exosystems characteristic of a given culture, subculture, or other broader social context, with particular reference to the developmentally instigative belief systems, resou rces, hazards, lifestyles, opportunity structures, life course options, and patters of social interchange that are embedded in each of these systems. The macrosystem may be thought of as a societal blueprint for a particular culture, subculture, or other broader social context (p. 228). Finally, the chronosystem includes the constancy or change over time of persons and the environment. Changes can be triggered by developmental changes, life events, or even socio historical experiences. Bronfenbrenner (1 989) illustrates this point by presenting the example of how people might change before or after a particular life experience or life transition. Ecological Model and Risk The use of the ecological model has been consistently applied to the study of risk in children and adolescents (Sameroff, 1985; Sameroff & Fiese, 2000) This approach acknowledges the many levels of var iables that interact with students which impact their academic achievement. Using the ecological perspective recognizes the complexity of interactions that occur at each level of the environment as well as within the child. It is the interplay of the c omplex interaction between the child and the environment over time that determines the outcome, with r isk and adaptation depend ing on the interaction of

PAGE 22

12 multiple systems. Wyman (2003) report ed research show ing and adaptability to adversity are context dependent, suggesting that the interaction of the ecological systems can produces diffe rent outcomes in different children. The use of the ecological model views factors as occurring on multiple levels within the social organization as well as multiple domains of child development (Sameroff, 2003) The study o f risk and protective factors allows for the interdependent relationships among different variables of risk and protection across multiple levels, exerting a reciprocal influence on one another (Yates, Egeland, & Sroufe, 2003) While intuitively it would make sense that the child be the center of attention in regards to academic achievement, research has borne out that environmental factors may be of even greater importa nce (Rutter, 2000; Sameroff, 1985) Ecological Model and School Social Work The major prevailing framework in school social work is the ecological approach, which serves as a base level for the assessment and delivery of student services. The National Association of Social Workers Standards for School Social Work Services (2002) listed in ecological perspective, focusing on the students, as well as their interactions in the school environment, at home, and in community settings. A functional approach to assessment enhances understanding of the purpose and effect of problematic behaviors an d provides information for developing Constable and Alvarez (2006) discussed the specialization of school social work in more detail and presented the idea that while school social work reflects the general ideas of social work, it demands its own set of expertise that are geared to this specific fi eld. As school social workers develop

PAGE 23

13 specialized skills and continue to work in a unique environment, there is a clear need for social workers to develop and use an effective school social work specific model. With has to be built on empirically based theory in order (Johns son & Svensson, 2005, p. 431) this study will look at developing a school social work specific model for predicting academic achievement Dupper (2003) introduced the concept of the ecological perspective as an organizing framework to view school soc ial work practice. The benefits of using this systems view are presented as being able to be dually student and system focused. C onnect ing the person to the environment takes into account multiple levels of interaction and the impact of a network of soci al and interpersonal influences E ach level of circumstances within the context of the environment. Garrett (2007b) looked at the practical application of this concept and sugge sted school social workers should assess how the student impacts the environment and vice versa. The N eed for a M ore Comprehensive M odel While it is clear t he ecological model takes into account numerous factors which have an effect on an individual person (Belsky, 1994; Bronfenbrenner, 1989) a lack of uniform practice among school social workers exists due to the complexity of ecologic al theory (Clancy, 1995) School social workers must work through all levels of the system to enact change, requiring social workers to have complete vision of the micro, m es o, m acro, and chrono systems Interventions may need to be applied at the cultural level or even at the social institution itself. This wide range of possible interactions provides unique challenges for social workers and encourages the expansion of the way social

PAGE 24

14 work practices are see n. In the creation of a new school social work mod el for predicting academic risk the ecological perspective serves as the base foundation. This layer represents a hypothesized, theoretical approach looking at each system an d the interaction between the levels of the system and individual factors. However, to continue moving forward, each factor and interaction needs to be broken down into the most detail possible. Many current illustrations of this model are too broad to be useful in providing an accurate assessment of the student in the environment (Lucio, 2006; Wakefield, 1996a; 1996b) When looking at these domains from a broader ecol ogical perspective, Bronfenbrenner (1979) described development as being influenced by the family, school, peers and neighborhood. These wide ranging areas are often included i n ecological models, but not in enough detail to be applied specifically and efficiently to school social work. The ecological perspective and domain s as mentioned by previous authors are d escribed briefly and as large ranging concepts (Garrett, 2007a; Richman et al., 2004) and is illustrated in Figure 1 in Appendix A Describing school social worker as an (Germain, 2006, p. 36) is accurate, but provides little detail which is useful for applying this model to social work practice. A more functional approach would be to establish each of the factors making up the larger domains, which provides a more comprehensive and applicable model. For example, rather than simply noting school factors impact achievement, including teacher involvement as a factor under the school domain allows social workers to focus on areas that are more precise and identifiable. The goal of this research will be to present this

PAGE 25

15 model using a more detailed examination of the factors that have been identified through the literature as having a significant relationship with academic ach ievement. This entails looking specifically at one domain school factors examining each factor and then integrating them into a larger ecologi cal model of school social work. This model will focus specifically on school related factors which impact academic achievement. Other researchers have looked at components of this model, but have not drawn all of the factors under a comprehensive model of school factors that predict academic achievement Cumulative Risk Children often face numerous risks in their lives as well as recurring stressors. According to Sameroff, Morrison Gutman and Peck (20 03) focusing on a single risk each factor makes the ecological model a complex process which impacts the course of academic achievement. Masten and Coatsworth (1998) found risk factors tend to cluster in individuals and that it is common for children who are experiencing one risk factor to also have other risk experien (1987) p revious research with psychiatric disorders showed that no single risk factor was identified as increasing the likelihood of developing a disorder However, when any two stressors occurred together, the risk level went up four fold The relative risk of a single factor alone is much smaller than the cumulative risk experienc e. Wachs (2000) is one of many researchers who have determined that no single risk factor is sufficient enough to explain differences in outcomes. Many others have reiterated this sentiment and continued to report that n o one single risk factor by itself

PAGE 26

16 caused the detriment or success of student outcomes (Mitchell, Bee, Hammond, & Barnard, 1985; Sameroff & Fiese, 2000) Rather, it is the cumulative effect of factors that determines the impact on outcomes When looking at any one single risk factor, such as poverty or living in a single parent home, the impact of these variables is minimal when put into an ecological framework individually While each single variable may be significant, it is the accumul ation of multiple negative influences that is the distinguishing focus of high risk groups. factors in which these individuals and families are embedded (Sameroff & Fiese, 2000, p. 141) that is most important when looking at outcomes. Therefore, it is t he number of risk factors, rather than the nature, that appears to be the determinant of outco mes When looking to maximize child outcomes, Luthar and Zelazo (2003) pointed out considered together explain more variability in ou tcomes than do any considered (p. 514). This sentiment was echoed in the work of Garmezy and Masten (1994) who mentioned that risk factors often co occur with other risk factors. Often risk is experienced as se quence of events, rather than a single experience. Understanding this has allowed researchers to move to a cu mulative risk factor approach. The resulting impact of risk and protective variables is the interplay between multiple psychosocial and genetic f actors. Rutter (2000) found that with few exceptions the risks for negative ex planation for individual differences in response to psychosocial risk concerns the number of risk factors involved and the duration of t he individuals exposure to them (p.

PAGE 27

17 670) Overall, children who fair better have been exposed to fewer risk factors an d for a shorter period of time. Coie, Miller Johnson, and Bagwell (2000) found that many disorders have multiple risk factors, signifying there are multiple paths to the same outcome. Other researchers have come t o a similar conclusion that different combinations of risk factors can lead to similar outcomes in children, strengthening support for the use of a cumulative risk approach (Deater Deckard, Bates, & Pettit, 1998; Sameroff, 1985) Large individual differences among high risk children create differing responses to the specific risks in the environment, making it difficult to ide ntify single specific risk factors that would apply in the context of each situation. Empirical Support for the Cumulative Risk Model Numerous studies have been done that support the use of the cumulative risk factor approach when determining the risk of specific outcomes. The Rochester Longitudinal Study looked at ten environmental factors with each factor shown to be individually related to poorer outcomes (Sameroff, 1985; Sameroff, Seifer, Zax, & Barocas, 1987) However, no single factor related to positive or negative outcomes. The same outcomes were the result of different combinations of risk factors, allowi ng the authors to conclude that the number of risk factors present was the most important determinant of outcomes, not the magnitude of each individual risk factor. While each specific risk factor demonstrated a medium effect size allowing for group compa risons, they were not large enough individually to indicate if specific individuals would experience negative outcomes. However, creating a multiple risk score using the total

PAGE 28

18 number or risks, they found major differences on measures of mental health and intelligence. Another study by Furstenberg, Eccles, Elder and Sameroff (1999) in Philadelphia used twenty two variables selected across the entire ecological spectrum from the microsystem (parent child interaction) to the macrosystem (school climate) and also examined the impact across multiple outcomes. These resea rchers looked at psychological adjustment, self competence, problem behaviors, activity involvement, and academic performance. As with other studies, as the number of risk factors increased outcomes declined (Sameroff & Fiese, 2000) An odds ratio analysis found that academic performance was impacted the most with negative outcomes in the low risk group at 7%, which increased to 45% for the high risk group. Looking specifically within the educational context, children who were abused and neglected had lower cognitive test scores, grades, and behavioral functioning (Alaimo, Olson, & Frongillo Jr, 2001; Crozier & Barth, 2005; Kerr et al., 2000) In each of these studies as the number of risk factors increased, achievement related outcom es decreased. More specifically, Marchant, Paulson and Rothlisbert (2001) studied combinations of factors that led to more positive achievement and grades. They also concluded that no one factor and no single contextual variable was more influential than any other in predicting achievement. The same conclusions were drawn by Gutman, Sameroff, and Eccles (2002) who found that adolescents had lower grades, increased absences, and lower standardized test scores as their exposure to risk factors increased. Finally, these findings were replicated in a later study by Gutman, Sameroff, and Cole (2003) who tracked the impact of risk factors from the 1 st to 12 th grade. As students

PAGE 29

19 progressed from the 1 st grade to the 12 th g rade, it was discovered that high risk students outcomes decreased more than their low risk peers Additive Model P rotective factors ha ve the ability to counterbalance adversity in a model in which assets outweigh risks Researchers have called for a mod el which includes assets, competence and protective processes along with the traditional measures of risk factors, symptoms, problems and risk producing processes (Luthar, 1991; Masten & Powell, 2003) Sameroff (2003) argue d rather than seeing protective and risk factors as competing, they should be viewed in the context of ad ditive contributors to the positive and negative outcomes for children. It is not any single factor that is responsible for outcomes T here appears to be little difference between the influen ce of risk and protective factors. The more protective factors present, the better the outcome. Conversely, the more risk factors present, the worse the outcome. Not all children who experience risk factors end up displaying academic or social problems (Burchinal, Roberts, Zeisel, Hannon, & Hooper, 2006) T his is why it becomes vital to also include promotive factors in any model Often, f actors can be a risk factor or a protective factor depending on where they fall on the continuum. For example, lower maternal education is a risk factor, while higher mat ernal education can be a protective factor. Since many of the same environmental factors can impact students through risk factors also act through protective factors, Brooks (2006) suggest ed that including protective factors in the framework provides opportunities to recognize factors that promote positive youth development and prevent negative outcomes.

PAGE 30

20 Sameroff (1985) showed that when a n additive model was used for promotive factors, the results mirrored the cumulative risk outcomes. As the number of promotive factors increased, positive outcomes also increased. Families with a higher number of promotive factors did substantially bette r than those with fewer promotive factors. Just as single individual risk factors carry a small risk when occurring in isolation, protective factors viewed alone are also likely to have a very small impact (Rutter, 2000) Rather it is the cumulative effect of risk and protective factors which is most important in this approach. Research has shown that outcomes are more positive when not relying on removing stress an d adversity, but rather helping recognize and enhance protective factors (Werner, 2000) Condly (2006) stated resilience is essential to the design and implementation of policies and programs that attempt to redress some of the effects that communit y violence, family discord and abuse, and poverty and minority status can have on children Mullis (2003) sum med up the sentiment of the additive model by pointing out that deficiencies in some areas can be countered by strengths in other areas. A cademic performance is the result of a complex mix of individual characteristics and social influences. Empirical Support for the Additive Model Numerous studie s lend support for the impact of protective factors and make a strong case for their inclusion in any model. Gutman and Midgley (2000) looked at African American students and found that those students who had two protective factors had a significantly higher grade point average (GPA) than those who only had one factor

PAGE 31

21 present. While in some studies an absent father was shown to decrease achievement outcomes (Jones, 2004) Menning (2006) found that children whose father was involved were less likely to fail school. In this study as father involvement increased, so did achievement. This suggests that protective factors can influence outcomes over time and should be included in any model. Prelow and Loukas (2003) found that as they added positive factors to their model, problem behaviors became non significant. T he relative risk was reduced as more protective factors were present, implying that protective factors could offset the effects of risk. This research continues to lend credence to the position that positive factors may mitigate relationships between cumulative risk and outcomes. Even though youths might be expos ed to certain risks, these results can be greatly reduced through the impact of protective factors. Gutman, Sameroff, and Eccles (2002) found in their study that achievement Risk and Protective Factors Any assessment and diagnosis must fo lives as well as those factors that serve a protective function. There is a delicate balance between risk and protective factors that can shift over time. Werner (2000) reported that in studies of children over time some began by doing well and deteriorated over time, while others seemed to grow into competent adults despite an early high risk status. Any time in the life course, particularly at times of major transition there may be a rebound toward positive outcomes.

PAGE 32

22 Risk and protective factors may not be so different in that factors that produce maladaptive variations are no different in kind and perhaps degree from those that produce adaptive variations (Sameroff, 2003) The first step is understanding that risk factors involve estimating probabilities more than finding risks. Fraser (2004) illustrated this point using the interaction between school failure and poverty. While poverty has been linked to poor academic outcomes, not all children who come from p oor families will fail school. Some researchers have come to find that ofte n protective and risk factors lie on the extreme ends of the same continuum (good or poor health) some factors only create disorders (born to a young mother), others create only goo d outcomes (taking music lessons) while for other factors the effect can be curvi lin ear in nature with the maximum benefit lying in the middle rather than on the e xtremes (Luthar & Zelazo, 2003) Risk and competence indicators are more similar than different. While risk indices are constructs associated with negative outcomes, competence indicators are those represe nting positive outcomes. For example, the same outcome can be described through competence by looking at the presence of health or risk when examining the absence of disease. Finn (2006) used the example of academic engagemen t to illustrate behavioral risk factor when students (p. 10). Simply looking at risk and protective factors might not be enough to give an a ccurate picture of what is happening Fraser, Kirby, and Smokowski (2004) found that c ontextual effects are conditions within the environment that affect vulnerability. More specificall y, these include interactions within a social and environmental context that creates or maintains poor or positive functioning. Contextual effects often include being

PAGE 33

23 part of a specific group or classroom, such as living within a neighborhood that promo tes academic competence No matter what the definitions used, Sameroff (2003) argued that i t is important from a practical and intervention perspective to help identify shortcuts where definitions of risk and pr otective factors can be useful despite any theoretical simplification Risk Factors Risk factors were first termed for use in epidemiological research (Costello & Agnold, 2000 ) W yman (2003) define d risk factors as how much a child and their family were exposed to psychosocial adversity. Luthar and Zelazo (2003) have said factors can be considered risk (p. 514). A ccording to Sameroff (2003) in order to identify risk factors, two criteria must be met. The first is that the variable must correlate with one of the selected outcome variables and secondly, those families who experie nced the risk factor perform significantly worse than families without that environmental risk. For the purposes of this dissertation risk will be defined according to the definition set forth by Fraser, Kirby and Smokowski (2004) who define d that incr e ase the chances for harm, or more specifically, influences that increase the probability of onset, digression to a more serious state, or main tenance of a p roblem Protective Factors T here has been some confusion of terminology in relation to protective factors among resiliency researchers (Luthar, Cicchetti, & Becker, 2000; Rutter, 2000) Some authors use the term pr otective factors to describe all interactions among factors which

PAGE 34

24 have a positive impact on desirable outcomes. This has been contrasted by those who suggest protective factor refers only to those factors where a benefit occurs in the presence of risk, bu t not in the absence. R utter (2000) says that it is essential to demonstrate that a child has experienced an environmental risk that carries with it an increased risk for negative outcomes. In addition, in order to study protective factors there must be prior demonstration that the individuals concerned have in fact experienced a significant risk. Rutter (1987) define d protective factors as moderators of risk and adver sity that enhance good outcomes, while Masten and Po well (2003) described p rotective factors as those that offer competence under adverse conditions. Sameroff, Gutman and Peck (2003) suggest ed that using this terminology of protective factor can be problematic as it implies that there is a shielding nature to the risk factor, which may be true for some constructs but not all. In these cases, Sameroff and Fiese (2000) suggest using the term promotive factor as it more accurately reflects the nature of the interaction, which indicates that the benefit is independent of risk. Others have even suggested the term resource factor be used when a factor is equally as beneficial for those exposed and not exposed to risk (Ong, Phinney, & Dennis, 2006) Given the wide range of definitions for positive factors, including protective, promotive, and resource factors, this dissertation will use the suggestion of Luthar, Cichetti and Becker (2000) who used the term pr otective factor to refer to any competency enhancing factors. This definition will include promotive and protective factors under the same definition, w ith the final definition coming from Fraser (2004)

PAGE 35

25 who define d l and internal resources that promote (p. 5). Model Development and T esting The central goal of this study is to develop a model which identifies students who a re at risk of academic failure, and which school related factors are predictive of this outcome. In order to examine this relationship to academic failure (outcome), each of the individual factors which contribute to academic failure (risk factors) and those that promote academic achievement (protective factors) must first be revealed and defined. This section of the dissertation will discuss how the outcome measure was selected, as well as each individual factor. The final section of this chapter will discuss methodological issues regarding the design and testing of the model. Outcome V ariable: Grade P oint A verage (GPA) Grade point average will be used as the primar y measure of academic achievement. Beacon and Bean (2006) looked at measures of GPA and examined its use in research studies. These researchers not only examined the validity and reliability of GPA use, but also which measures of GPA were most predictive of curr ent and future success. Overall, they found when comparing three methods of GPA calculation, cumulative semester or only core courses. Often measures other than official school records are used either for convenience or due to the difficulty that is encountered when trying to obtain school transcripts. These measures can include student, parent, or teacher reported grades, test scores, and abilities. Given the reliance of numer ous studies on this methodology, Kuncel, Crede,

PAGE 36

26 and Thomas (2005) performed a meta analysis specifically looking at the validity of self report grades. These authors exam ined 37 samples, which involved 60,926 subjects, using a psychometric meta analytic method. They also reported that t his statistical method has the ability to account for variability across effect sizes due to random errors. Their analysis found that hig h school self report GPAs ( r =.82) were less accurate than college GPAs ( r = .90), and that White student reports were higher ( r = .80) than non white students ( r = .66). They also found that some students had a tendency to over report their GPA (12.2%), while a smaller number under reported their grades (3.3%). S tudent s with lower levels of school performance were more likely to inaccurately report grades than students with higher GPAs. The results of this meta analysis suggest self report grades can be a reasonably good predictor of grades, but should be used with caution and only when it is not possible to obtain school transcripts. Domains An extensive literature review was performed to identify the broad domains and individual risk/protective factors related to academic achievement. Numerous authors have looked at the different domains that have been used to classify risk and protective fa ctors. When looking at these domains from a broad ecological perspective, Bronfenbrenner (1979) proposed that development is influenced by the family, school, peers and neighbo rhood. These domains have been commonly referenced throughout the literature by numerous other authors, with others adding student characteristics as an additional domain (Brooks, 2006; Luthar & Zelazo, 2003; Yates et al., 2003) This study will focus exclusively on the school domain, and the factors classified under this domain. Recently the Response to Intervention model has been pushed to the front of the school

PAGE 37

27 social work agenda as changes in the law have opened the door for implementation of evidence based curriculum and interventions (Pub. L. No. 107 110) Some states have begun to look at this as a statewide model, which demands school social workers be well versed in integrating this into current pract ices (Bureau of Exceptional Education and Studen t Services, 2006; Colorado Department of Education, 2005) It is within this context that the need for research within the school domain becomes a vital need for school social work. Being able to identify school related factors which impact achievemen t is essential when operating within a school setting. Knowing which risk at universal, group, and individual levels. The process for determining which factors b elong under the school domain will be discussed in detail in the following sections. School Related F actors To determine which factors should be included in this model a literature search was completed to identify factors that have been associated with academic achievement through previous research. This was accomplished by searching the abstracts in PSYCINFO, Social Work Abstracts Plus, and ERIC (Educational Resources Information Center), using the search terms academic achievement and educational outc omes An initial se arch revealed a total of 105,816 articles, whic h was further reduced to 20,717 when looking at articles only from the years 2000 through 2008 (See Table 1 in Appendix A ). Each abstract was reviewed and relevant articles whic h addressed factors associated with academic achieveme nt were examined in more detail. The initial examination of the literature revealed a total of 11 8 distinct factors that have been previously related to academic achievement through other studies (see Table 2

PAGE 38

28 in A ppendix A ) A number of factors exist under each domain and a concept mapping activity was used to identify the specific domains to which they should be placed. Once the factors were classified according to their broad categories, the individual factors for the school domain were examined in more detail. Concept Mapping The use of concept mapping has been developed through the work of Trochim (1989a) and is a structuralized process designed to develop a conceptualized framework for evaluation and planning. This method has been applied to curriculum development (Keith, 1989) psychiatric rehabilitation (Shern, Trochim, & LaComb, 1995) mental health (Johnson, Biegel, & Shafran, 2000) and students who spoke English as a second language (Chularut & DeBacker, 2004) T he goal of concept mapping is to organize complex and diverse ideas into an understandable and coherent framework (Trochim, data into a pictorial form that displays the in (Johnson, Biegel, & Shafran2000, p. 67) Trochim (1989a) outlines concept mapping through a six step process which involves preparation, generation of statements, structuring of statements, representation of statements, interpretation of maps, and utilization of maps. The first step, preparation, includes selecting the participants and deciding the specific question s Trochim believes this to be one of the most important tasks and should include a wide variet y of relevant people. The second step, generation of statements, involves the generation of statements which represent the entire conceptual domain of the topic of interest. Traditionally this is done through a brainstorming process with the participants who generate a list of

PAGE 39

29 statements that represent the broader conceptual domain. The third step in the concept mapping process involves structuring the statements. The goal is to have each factor rated on one dimension and sort out how the factors are related to each other. In some cases the statements are sorted on single cards and placed into piles according to how they make sense to the participants. The resulting outcome is based upon how often the factors were related into the same domain. The fourth step, representation of statements, uses the application of multidimensional scaling and hierarchical cluster analysis. Statements are located as points on a map, where statements closer to each other have been sorted together more frequently and t hose more distant were sorted together less frequently. These statements are then grouped into clusters on the map, representing conceptual groupings from the original statements. The final task in step four is to overlay the average ratings either by po int or cluster. Johnson, Biegel, and Shafran (2000) any two points in this two dimensional plot reflects the degree to which the statements represented by the points were seen as conceptually similar by the group, and thu s sorted The fifth step, interpretation of maps, combines the maps generated in step four with the statements created in step one. The clusters are examined and discussed by the focus group par ticipants, and are labeled by a descriptive phrase which captures the nature of the cluster. Participants are also asked whether the conceptual mapping makes sense and then the clusters are named and identified. The sixth and final step, utilization, det ermines how the maps will be used. This goes back to the initial step and combines the purpose of the mapping activity with the interpretations revealed in step five.

PAGE 40

30 Concept M apping A ctivity Drawing on the work of Trochim (1989a) a concept mapping activity was employed to classify the factors that wer e identified through the literature search into specific domains. This activity involved adapting the concept mapping process to suit the specific demands of the project. Previous research has shown modifications to the methodology can be done and still maintain the integrity of the outcomes (Johnson et al., 2000; Trochim, 1989b) Eighty school social workers were given the opportu nity to complete the activity, and fifty eight provided feedback (72.5%). Each of the social workers selected were employed as a school social worker within one west central Florida C ounty. These participants were selected because they are practicing sch ool social workers, and have a diversity of experience in all aspects of school social work including itinerant, school based, teen parent, drop out prevention, and emotional behavioral disorders. It is also important that a school social work model for predicting academic achievement be developed by and represents the views of school social workers. Generation of statements is considered the second step and was modified to reflect that factors were selected through an extensive review of the literature This review identified 11 8 factors that have been associated with academic achievement through previous research. This type of adaptation has been used previously when the statements are already known and has the advantage of applying an already impli cit structure to the conceptual domain directly without asking p eople to generate statements (Trochim, 1989a) The third step, structuring the statements, was completed by providing each social worker with a form listing the identified factors and having each of

PAGE 41

31 the social workers classify the factors according to one of the 5 previously identified domains of student characteristics, family factors, peers and friends, school factors, and neighborhood or community factors ( see Appendix B ). The fourth step involves using multi dimensional scaling to determine which fac tors cluster together and then place the factors into the appropriate domain. In this case, two factors were redundant and removed from the analysis. Economic status appeared in two places in the form and one was removed from the analysis. The remaining data was then entered into SPSS and analyzed for missing data and initial breakdown of factors into domains, with each factor being give a score of 1 for child, 2 for family, 3 for peers, 4 for school, or 5 for community, reflecting the proximal relations hip to the child. The results indicate there were no missing data and that 69 of the items achieved an agreement of 75% or higher (see T able 3 in Appendix A ). Factors that had an agreement of higher than 75% were placed into their corresponding domain. The remaining 48 factors, along with the 5 from each domain with the highest agreement percentage, were entered into the multi dimensional scaling (MDS) model. This was done to accommodate the 100 factor limit of SPSS, and still retain the distance plot ting which is central to MDS. Figure 1 in Appendix A illustrates the resulting plotted points, which were grouped into each of the five domains according to their location. The fifth step, interpretation of the map, included an analysis of the plot. Th is revealed the factors were clustered around the 5 central domains. Each of the factors was then placed into the appropriate category. This resulted in 26 child related factors, 48 family related factors, 8 peer related factors, 24 school related factor s, and 14 neighborhood related factors. Table 4 provides a detailed illustration of each domain.

PAGE 42

32 The sixth and final step, use of the data, relates to the goal of this study which is to develop a school social work specific model, using school related fa ctors and the relationship to academic achievement. This will be completed throughout the rest of this study. School F actors The primary consideration of factors to be included in this study was identifying items through a conceptual base, as well as th rough previous studies which showed a casual relationship to academic achievement. Initially, an extensive literature review was completed to reveal which factors were related to academic achievement through previous research This was followed by the cl assification of factors into 5 domains by practicing school social workers The focus of the model is school related factors, and t he concept mappin g activity identified 2 3 potential items for inclusion in the model and analysis. The factors identified by the school social workers were academic engagement, academic expectations, academic self efficacy, attendance, class size, educational support, family type (school), grade retention, homework, music instruction, sc hool behaviors, school belonging, school district size, school minority rates, school quality, school relevance, school safety, school SES, school size, school mobility, supportive school environment, teacher relationships, and teacher support. This numb er was reduced to 16 factors by removing class size, school SES, school district size, school minority rates, school quality, school size, and school family type. These 6 factors each present a valid reason for excluding them in the analysis. Class size has been mandated by state law to be less than 25 per class on a school wide average for all students (Constitution of the State of Florida. Article IX. Section 1.7., 2007) school district size

PAGE 43

33 school SES, school minority rates, school quality and school size would also be the same for each participant which crea tes zero variability in each of the se factors The final factor, school family type, is not available through school records and would be beyond the student s knowledge Each of the remaining school related risk and protective factors identified through the concept mapping activity will be discussed in more detail in the following section. Academic E ngagement F inn (2006) note d student engagement has been found to be one of the most robust predictors of student achievement and behavior in school and state d who participate actively in school and the classroom, and who identify with school, increase the likelihood of Klem and Connell (2004) reported that students with high levels of engagement demonstrated better grades, attendance, and graduation rates than those students with low engagement. These researchers performed a threshold analysis to identify those students doing well (optimal levels) and those students not doing well ( risk levels) to determine a tipping point where the chances of success increase significantly. Middle school students who had higher levels of engagement as rated by their teachers were 75% more likely to be successful on measures of attendance and achiev ement Conversely, those with low levels of engagement had a lower likelihood of success. This study used multiple measures of student engagement, including student and teacher reports, and had a large sample size of 3,300 students aged 7 to 15. Additi onally the measures of engagement that were used all However, among

PAGE 44

34 elementary aged students the free/reduced lunch rate was 85%, which was considerably higher than the 58% reported for middle school students. An examination of African American students aged 12 19 and their mothers yielded similar findings (Sirin & Rogers Sirin, 2004) Looking to expand the knowledge base of middle class African American students, a regression analysis revealed engagement was related to academic performance. I t should be noted that the participants in this study were middle class, primarily African American, over half of the parents had a college education and the study was based on a cross sectional design While these factors have t he potential to limit the generalizability of this study, when added to previous studies this suggests there academic engagement has a broad ranging impact of a cross different ethnic and economic backgrounds. Academic E xpectations (St udent) In a study of academic achievement it was found that students grade goals were directly involved in final grade outcomes (Sirin & Rogers Sirin, 2004) Zimmerman, Bandura, and Martinez Pons (1992) asked high s chool students to identify the lowest grad e they would find satisfying. Those who reported higher grade goals ultimately ear ned higher grades in school. Abu Hilal (2000) also studied a academic aspiration using s tructural e quation m odeling and found a positive relationship question in which students were asked how far they expected to go in school. T he sample consisted of 280 high school students, which included 121 boys and 159 girls, though there was no other information regarding race or other demographic factors which may have had an influence on the outcomes.

PAGE 45

35 Academic S elf E fficacy Bandura (1986) defined self efficacy as beliefs people have about whether or not they can successfully complete a task. W ork by Zimmerman, Bandura, and Martinez Pons (1992) revealed academic self efficacy had both direct and indirect effe cts on final grades in school. This study consisted of 102 students in the 9 th and 10 th grades, with a path analysis revealing both grade goals and academic self efficacy had direct impacts on final grades. The analysis also took into account the school and class membership, eliminating some of the confounding variables. Numerous other studies have supported the conclusion that perceived academic competence was predictive of current and future grades (Gonzalez Pienda, Carlos Nunez, Gonzalez Pumariega, Alvarez, Roces, & Garcia, 2002; Obach, 2003) The results are consistent even when looking at ethnic minorities such as Latino students (Alva & de Los Reyes, 1999) and African American students (Saunders Davis, Williams, & Williams, 2004) or controlling for prior achievement (Tavani & Losh, 2003) This suggests that building students beliefs about their academic self (Davis, Saunders, Sharon, Miller Cribbs, Williams, & Wexler, 2003) Attendance Attendance in school is directly re lated to academic achievement and also linked with completion of school, test scores, and grades (Dunn, Kadane, & Garrow, 2003; Gambone, Klem, & Connell, 2002; Powell & Arriola, 2003) Crean, Hightower and Allen (2001) studied teenage parents and found higher rates of attendance were associated with higher graduation rates. While they did not target specifically cognitive abilities in the program, they did require the mothers to attend 80% of the time.

PAGE 46

36 Interestingly, the percentage of courses passed and changes due to the birth of the child did not have an impact on graduation. They caution that using pass/fail as dichotomous variable may not have been sen sitive enough to notice smaller changes in academic outcomes. Klem and Connell (2004) used a threshold analysis and identified an attendance rate of above 80% as being the point at which secondary students are likely to be more successful in school. For elementary students, this percentage was even higher at 90%. A threshold analysis differs from traditional methods in that threshold levels identify point or threshold Educational S upport Having a family that was supportive and understanding of the need to study and complete school work has been related to increased grade point averages (Ong et al., 2006) While this study was primarily focused on Latino families, Powers, Bowen, and Rose (2005) examined a national non probability sample of 10, 344 students across six states and foun d that parents educational support was related to self report grades. This factor continued to be significant even though it was analyzed with 21 other main factors. A similar study by Bowen Bowen, and Ware (2002) looked at another national sample of 1,757 students and found the same results. In both studies, support was measured by asking students if an adult in the household helped with their homework, talked about school with them, discussed their future with them encouraged them to do well in school, or limited television time and time out with f riends on school nights. This conclusion has been maintained by Henderson and Mapp (2002) who found w hen students report

PAGE 47

37 feeling support from both home and school, they tend to do better in school Gutman and McLoyd (2000) also reached the same conclusion when they found that f amilies of high achieving students were more likely to have conversations centered around endeavors Grade R etention While retaining students in early grades is design ed to prevent future failures, a review of the literature found that grade is one of the most powerful predictors of later school withdrawal (Jimerson, Anderson, & Whipple, 2002, p. 452) Jimerson, Carlson, Rotert, Egeland and Sroufe (1997) followed the same cohort of children from Kindergarten through age 16. They found when compared to a similarly low achieving group that was promoted and not retained, standardize d test scores at the end of the following year were the same despite the retained group getting an extra year of the material. Additionally, following up at age 16 revealed no significant differences between those students who were retained and those who were not on standardized test measures. This was a significant study because it compared outcomes for a group of low performing retained students low performing not retain ed students and a group of non retain ed regular performing students. A similar study by Jimerson (1999) followed a cohort of students for 21 years and found that the retained group was 20 25% more likely to drop out of schoo l than a comparable low achieving, but not retained, group. Moller Stearns, Blau, and Land (2006) looked at the growth curve trajectory for retained and non retained students and found the growth curve trajectories were similar for early retentions (prior to 2 nd grade) and later retention (2 nd through 7 th grade). In fact, when

PAGE 48

38 adding retention to the model the percentage of variance accounted for increased by 8%. The results of this study suggest that predictions of achievement would be substantially more accur ate if a s tudent s previous pro motion patterns through school are included. Homework Homework factors have been shown to have a connection with academic achievement i n grades 6 12 Cooper, Lindsay, Nye, and Greathouse (1998) found percentage of homework completed was associated with higher rates of graduation and better grades. The use of percentage completed more accurately reflects the impact of homework by recognizing homework assigned and homework time differ by school, teacher, and subject. A path analysis found a positive relationship between homework completion and achie vement even when controlling for amount of homework assigned, grade level, and subject matter. The authors did note the 35% response rate was lower than previous studies on homework, but speculated this was due to a longer survey that parents, students and teachers all had to complete the survey in order to be included in the analysis. Little is known about those parents, students and teachers who did not respond. While this study did not find the number of hours spent on homework to be significant, other researchers have found a positive relationship between those variables (Rumberger & Palardy, 2005) Music I nstruction In a study of 15, 431 public school students, instrument playing students outperformed their non instrument playing peers in every subject and every grade level (Fitzpatrick, 2006) Students were matched by SES level and classified as instrumental if

PAGE 49

39 they had enrolled in band, orchestra or jazz during the school year. T tests were run for each group of students who were divided into groups based on their SES level, and it was found the results held across all levels of socioeconomic status. However it should be noted on e of the major limitations of this study was a large difference in sample size between instrument playing ( n = 915) and non instrument playing ( n = 14,516) students. Schellenberg (2006) looked at the association between music lessons and academic outcomes in several studies. Both the long term effects of music lessons and the impact music lessons had on multiple measures of academic success were examined. The first study involved looking at the impact of formal exposure to music in childhood and academic performance during undergradua te education. This research suggests that early music instruction has a small, but broad and positive impact on academic ability in high school. A second study by Schellenberg (2004) used an analysis of variance to look at the IQs of students who received 36 weeks of instruction in music. After randomly assigning students to mus ic instruction (keyboard or voice) and non music instruction groups, it was found that students who receiv ed music instruction displayed significant increases in IQ. These increases were determined to be a medium effect size. School B ehaviors Student b ehaviors in school can be one of the strongest predictors of academic outcomes and can be positive or negative in nature (Prelow & Loukas, 2003) Positive behaviors include doing schoolwork, atten ding class, and following school rules, while negative behaviors might involve copying homework, cheating on a test, getting sent to the office, misbehaving in class, not follow ing teacher directions, or skipping school (Bryant, Schulenberg, Bachman, O'Malley, & Johnston, 2000; Marks, 2000; Powers et

PAGE 50

40 al., 2005) One nationally stratified study of 24,599 stude nts looked at multiple school behavior related variables which includ ed skipping school, referrals for school behaviors, or fighting in school and found these behaviors were related to test scores in math, science, reading, and social studies (Mullis et al., 2003) School behaviors had the largest direct impact on self reported grades, but grades were also influenced by parental resources, involvement in activities, connection to their school. Research by (2000) examined a total of 3,056 students in 8 th 10 th and 12 th gra de using a s tructural e quation m odeling approach to look at school misbehavior and academic ach ievement. The resulting model revealed that s chool misbehavior interacted direct ly with academic achievement. School B elonging A sense of belonging to a school can be an important factor in the lives of students. A sense of belonging taps into inclusion, and respect and encouragement for participation (Sanchez, Colon, & Patricia, 2005, p. 622) Finn (2006) review of the literature revealed a membership, sometimes referred to as school belongingness, school bonding, or school connectedness, was related to school rel ated outcomes Anderman (2003) and Goodenow (1993) both looked at student connectedness and found when students felt a sense of belonging, in that they felt comfortable and respected in a particular school, they performed better academicall y. Zand and Thomson (2005) looked specifically at African American students and found school bonding also impacted the relationship with self reported grades. They also reported that in their sample of 174 students, sch ool bonding functioned as a mediator in the relationship between self w orth and academic

PAGE 51

41 achievement. While the sample was small and used self reported instruments, it does begin to illustrate the contextual components of school bonding. School M obility S tudents are more likely to learn about academic sub jects when they are in the same school Frequently, changing schools disrupts the ability to provide a cohesive course of study that builds on previously acquired concepts. Dunn, Kadane, and Garrow (2003) studied over 1 800 11 th grade students an d found that both school mobility and absences were negatively associated with academic achievement. They equated the cost of each move as having the same impact on academic achievement as fourteen days absent from school. Swanson and Schneider (1999) examined the e ffect of educational and residential mobility on a national sample of over 22,000 students. They examined students who move d to a new school but did not change addresses (movers), those who change d schools but did not move (changers), those who move d and change d schools at the same time (leavers) and those who remain at the same school (stayers). Using an O rdinary Least Squares regression they found moving or changing schools between grades 8 and 10 had little or no impact on mathematic achievement and b ehavior problems but changing after 10 th grade resulted in similar difficulties as those who dropped out. In addition, s tudents who had a greater number of school changes before the 8 th grade were significantly more likely to leave school between 8 th and 10 th grades than students who are non mobile School R elevance A study by Alpert and Dunham (1986) approached the problem of school drop out by looking at which factors are likely to keep kids in school. They studied 57 high school

PAGE 52

42 students who fit the profile of likely drop out s, yet remained in school. After school misbehavior, s chool relevance was found to be the sec ond leading predictor of keeping kids in school, which was determined by asking if finishing school would help students get the job they wanted. This tapped into the notion that schools providing the educational needs the students perceived as vital to their futures were more likely to keep children engaged in s chool. Davis, Johnson, Miller Cribbs, and Saunders (2002) looked at toward school social norms, perceived control, importance to completing scho o l and sel f esteem, which w ere all positively related to grade point average In this sample of 231 African American 9 th grade students, a hierarchical regression analysis revealed only attitude was directly related to GPA. Ultimately, those students who believed that school was important and a rewarding experience showed the highest grade point averages. School S afety S self report about feeling safe at school were related to grades in several studies. When defined as whether students defined t heir school by the level of crime, problem behavior, and bullying behavior school safety was significantly related to grades in a national sample students (Powers et al., 2005) R umberger and Palardy (2005) found the same result when they looked at the percentage of students feeling safe at the school and achievement related outcomes These results were consistent for math, reading, and science scores, and remained true even when considering school structure, teacher expectations, homework time, and rac ial composition of the school. Roscig n o (2000) looked at 1,239 students from 1 st through 8 th grade, who were a mix of White students (52.3%) African American students ( 31.8%), and Hispanic students (15.9%).

PAGE 53

43 Using standardized tests scores as a measure of achievement, the author utilized a multi step regression approa sex The findings indicate d s chool s with higher c rime rates, which looked at items such as gang activity, weapons on campus, robbery, theft and teacher assaults, were also linked to poorer achievement among students Supportive S chool E nvironment Magdol (1994) stressed that the general atmosphere of the school is an importan t having effective teachers, a flexible curriculum, and supportive administration. Bowen and Richman (2005) describe d a supportive learning climate as one in which students get a where the adults at school affirm and care about students, and where every student is valued. Marchant, Paulson and Rothlisberg (2001) looked at the school environment of middle school students as measured by school responsiveness and supportive social environment. Through a path analysis they found a supportive social environm ent at school was positively related to students reported grade s A mostly white sample and the use of self report measures limit the scope of the findings, but shows that this may play an important role in achievement. Teacher Support/R elationships Y who perceived their teachers as more supportive and caring also did better on measures of ac ademic performance and were more engaged in school (Klem & Connell, 2004; Powers et al., 2005; Woolley & Grogan Kaylor, 2006) Muller (2001) studied students who were at risk of drop ping out of high school and concluded that t he value of having a caring teacher may substantially mitigate the negative outcomes

PAGE 54

44 a ssociated with being at risk. Both early and current relationships with teachers have been shown to have an impact on later achievement. Examining early teacher relatio nships, Hamre and Pianta (2001) looked at 179 gi rls who had close relationships with their kindergarten teachers and found they h ad more positive work habits and fewer behavioral problems through elementary school. Following this group of girls through the 8 th grade revealed these students had higher better grades than those students who did not have a close relationship with their teachers. Negative child teacher relationships on the other hand were indicative of poorer academic outcomes, suggesting the qualit y of the rela tionship is an important factor in academic achievement. These results were consistent even when controlling for gender, ethnicity, IQ, and problem behaviors. Looking at current grades, Murray and Ma l m gren (2005) studied an intervention that linked teachers and African American high school students in a program to develop supportive teacher relationships During the 5 month intervention stage 48 students with emotional or behavioral problems and one of their teachers met weekly to discuss school related goals, strategies that would assist the students in meeting the goals, and reviewing progress toward meeting goals. At the end of the study time period, s tudents who were involved in the program reported higher grades in math, English, social studies and science, than the control group who had no increased interactions with their teachers. Klem and Connell (2004) looked at teacher support and found the i mpact on academic achievement may be partially mediated through s tudent engagement. Students who saw te achers as more caring, creating a well structured environment, and set ting high, clear and fair expectations were more likely to be engaged in school. Using a threshold analysis, t hey found that students experiencing higher levels of teacher support

PAGE 55

45 were three times more likely to be highly engaged in their education. Students who experienced low levels of teacher support were 68% more likely to be dise ngaged. Engagement in turn, has been shown to have a strong relationship with academic outcomes, suggesting the relationship has may have both an indirect and direct relationship with achievement. Limitations of Previous Research Other researchers have lo oked at components of this model, but have not drawn all of the factors under a comprehensive model. The previous research has been limited by the number of variables included in the study, the type of measures used, sample type, and sample size. While e ach of these factors creates problems with generalizability of the outcomes found, limitations are often a necessity in research. Zand and Thomson (2005) acknowledged these limitations when they noted t understandable and often necessary due to time constraints, limits in budget, access to records or specific populations, and even the nature of the research being conducted, they will be acknowledged and discussed. Number of variables Many studies have looked at only a few components of the overall mo del that is being developed. This is a necessary first step in analysis to begin looking at which factors impact achievement. However, as the model begins to grow it is important to combine all of the factors together in one model, as some of the factors may mediate the relationship of others, or even account for some of the impact of another factor. The

PAGE 56

46 most comprehensive research was the work of Powers, Bowen, and Rose (2005) who examined 21 factors which impact adolescents. Of these 21 factors, this includes only the 7 school related factors of school satisfaction, teacher support, school safety, home academic environment, school behavioral expectations, and parent educational support. Other researchers have examined several variables together inclu ding school behavior, school bonding, and cigarette use (Bryant et al., 2000) Zand and Thomson ( 2005) looked at the impact of self worth, school bonding, leadership, and independence, while Dunn, Kadane, and Garrow (2003) examined school mobility and absences together. Wh ile some researchers have examined numerous factors together, others have begun with only one or two factors. Klem and Connell (2004) looked at academic engagement and support together, while Zimmerman, Bandura, and Martinez Pons (1992) examined grade goa ls and self efficacy. Jimerson, Carlson, Rotert, Egeland and Sroufe (1997) looked at grade retention on achievement related outcomes, but looked at no other variables in their analysis. The limitation of not examining all of the factors together is also true in the research o n homework (Cooper et al., 1998) mu sic instruction (Fitzpatrick, 2006; Schellenberg, 2006) school behaviors (Mullis et al., 2003) attitudes toward school (Davis et al., 2002) and school safety (Rumberger & Palardy, 2005) Self Report G rades The use of self report grade outcomes is often used due to the ease of getting self report grades versus obtaining school records. School records are the most accurate measure of grade point average. Students and parents have a tendency to overesti mate a (Stone & May, 2002) ilities also tends to be inaccurate (Eckert, Dunn, Codding,

PAGE 57

47 Begeny, & Kleinmann, 2006) Numerous st udies have included self report grades as the outcome variable including research on educational support (Powers et al., 2005) school behaviors (Mullis et al., 2003) scho ol bonding (Zand & Thomson, 2005) and supportive relationships (Marchant et al., 2001) which may limit the accuracy of the findings. Analysis Tools Zand and Thomson (2005) examined school bonding, using single informant tools of moderate internal consistency reliabilities (.50 to .6 5 ). The reliability of the instruments was below the minimum standard of .70 set by Nunnally (1978) and DeVellis (2003) DeVellis suggested that a value below .60 is unacceptable, between .60 and .65 as undesirable, between .65 and .70 as minimally acceptable, between .70 and .80 as respectable, and between .80 and .90 as very good. Sample Population The population of the sample used can also limit the overall generalizability of the findings. Studies are often limited to using specific populations, either by design or location of the study. While this is an important step in looking at the contextual nature of specific factors (Richman et al., 2004) it is limiting to a larger scale model. African American students were the primary focus in research on academic engag ement (Sirin & Rogers Sirin, 2004 ) academic self efficacy (Saunders et al., 2004) school bonding (Zand & Thomson, 2005) and school relevance (Davis et al., 2002) Other populations that have gotten specific attention in research include teen parents with attendance (Tavani & Losh, 2003) Latino students with self efficacy (Alva & de Los Reyes, 1999) white students with supportive relat ionships (Marchant et al., 2001) and girls with teacher relationships (Hamre & Pianta, 2001) In addition to these studies, two studies

PAGE 58

48 (2000) work on school safety and Anderman (2003) who examined school bonding. Sample Size Several studies have also included small sample sizes in their analysis, which again is likely to limit the scope of the findings. When looking at academic expectations, Zimmerman, Bandura, and Martinez Pons (1992) examined the results of 102 high school students, while Albert and Dunham (1986) looked at 57 students in their exploratory study on school relevance. This is similar to the 48 students that were lo oked at in regards to teacher relationships by Murray and Mamlgren (2005) Finally, Fitzpatrick (2006) had a large sample size difference between music playing and non music playing student s in a study on music instruction. Summary Evans (1999) argue d uting to (p. 165). In a review of literatur e, Bruns, Moore, Hoover Stephan, Pruitt, and Weist (2005) note d that school interventions have the potential to impact emotional and behavioral problems which can prevent or ameliorate academic outcomes such as improved achievement, attendanc e or even school level outcomes. The first step in this process is to identify which factors impact achievement, followed by specific interventions for the selected factors. This dissertation aims to pinpoint which facto rs within the school domain impact academic achievement, creating an opportunity for school social workers to create more accurate assessments, which in turn guides the areas that need intervention.

PAGE 59

49 While many authors have looked at the ecological perspect ive as it relates to school social work, it is too often a wide framework which fails to adapt the perspective to school social work (Dupper, 2003; Garrett, 2007a; Lynn, McKay, & Atkins, 2003) avail able in most high schools career and academic counseling, mental health services, and a range of other problem oriented services (National Research Council, 2004, p. 156) There is already an abundance of research that reflects the nature of interactions between factors which are important in helping construct and expand a the oretical framework of school social work (Mehana & Reynolds, 2004; Teasley & Lee, 2006; Thomlison, 2004) While it is clear that much of the information is already available, the challenge for school social workers is to be able to bring all of this under one school social work model for predicting academic risk This process is built on a commitment to construct a strong new paradigm for school social work, thus adding to the ability to make acc urate and useful assessments Joining what we already know through previous research with an elaborated school social work specific ecological perspective will start to focus the research efforts in this part of the social work field This study will combine school related factors that have been selected into domains by school social workers, in order to develop a school social work specific m odel. These factors have been shown in various research studies to have an impact on achievement related outcomes. Figure 3 in Appendix A illustrates each of these factors and how they are measured. The current study will bring the school related factors into one model to determine which ones combine to best predict academic achievement. Any factors that do not impact

PAGE 60

50 achievement in the presence of other factors will be removed, revealing a school social work driven ecological mod el of school related factors.

PAGE 61

51 Chapter Three Methods P articipants Participants were 217 (44 males, 173 females) high school aged students ( M = 17.00 years, SD = 1.22), from three different school sites School A provided 197 students, school B had 10 students, and school C also had 10 students. The three schools were selected for the recruitment to ensure a wide variety of Grade Point Averages (GPAs). School A was a regular high school with a total e nrollment of 1,932 students, while school B (N = 231 ) and School C (N = 49) were both drop out prevention schools. Sixty two percent of the sample was White, 24% Hispanic, 9% Black, 3% Multi Racial, 2% Asian, and 1% American Indian, reflecting the diversit y of the sample across the three sites. Sixty seven percent reported they lived with two adults, while 22% lived with one adult, 8% lived in another family situation, and 3% reported living alone. In addition, 62% of the sample received free or reduced l unch. Finally, 11 th graders comprised 31% of the sample, followed by 12 th graders (26%), 10 th graders (24%), and 9 th graders (19%). All students were enrolled in the school district since the beginning of the school year to ensure that a cumulative GPA was available for analysis Table 5 in Appendix A shows a breakdown of demographic factors in aggregate, as well as by individual school site. Response Rates While letters were available to all students enrolled at each school, some were not sent home d ue to students not attending during the time frame letters were sent home. This reduced the potential sample from 2,212 students to 1,974. Of the 1,932 delivered

PAGE 62

52 to School A, 124 were returned as not being sent home. In School B, 97 of the 231 were retu rned unsent, while School C had 17 of 49 returned as not sent home. The overall response rate was 10.99%, though the response rates varied from school to school. School C had the highest participation rate with 31.25%, followed by School A (10.89%), and School B (7.46%). Research Design The design for this d issertation was a cross sectional survey design (Grinnell, Unrau, & Williams, 2005; Rubin & Babbie, 2004) McMurty (2005) reports cross on each student was surveyed once and the results were combined with school data collected at the end of the school year to create a complete dataset for each student. Controlled Factors There were several demographic variables that have been shown to impact academic achievement in previous studies and include individual socio economic status (SES), race, and gender (Eamon, 2002; Gorard, Rees, & Salisbury, 2001; Kellow & Jones, 2008; Ma & Klinge r, 2000; Marsh, Martin, & Cheng, 2008; Perie, Grigg, & Dion, 2005) In both the cumulative risk and additive risk models, several these factors were entered into the model first to control for their impact on the overall outcomes This was done to account for variance in overall GPA that may be related to a significant, but not school related, factor.

PAGE 63

53 Instrument Measures were included from various sources, including scales, subscales, and school records. A deta iled discussion of each measure will be provided in this section. The instrument and school records cover 1 dependent variable, 4 controlled factors, and 1 5 school related factors. These include cumulative GPA (dependent), SES (controlled), race (controlled), living si tuation (controlled), gender (controlled), academic engagement (school factor), academic expectations (school factor), academic self efficacy (school factor), attendance (school factor), educational support (school factor), grade retention (school factor), homework (school factor), music (school factor), school mobility (school factor), school safety (school factor), school behavior (school factor), school belonging (school factor), school relevance (school factor), and teacher suppo rt/relationship s (school factor) This totaled 20 factors for potential examination in this study. A complete version of the instrument can be seen in Appendix C Dependent V ariable Academic Achievement (school records) Cumulative g rade point average (GPA) was measured thro ugh school records. These records indicate d grade point averages on a scale of 0.00 (F) to 4.00 (A) and was Controlled Factors SES (school records and question 11) SES was recorded from school records through free or reduced lunch status In addition, poverty was also measured through material hardship as asked by Beverly (2001) which ask ed dichotomous (yes/no) questions about food insufficiency, eviction,

PAGE 64

54 utility disconnection, telephone disconnection, clothing needs, and l ack of school supplies. The internal reliability of the material hardship scale was measured for this data set using the Kuder Richardson formula (KR 20) as .679. Race (school records) Race was obtained from school records and was classified as White, Bl ack /African American Hispanic /Latino Asian or Pacific Islander, American Indian or Alaskan Native and Multiracial Gender (school records) Gender was ascertained from school record and is listed as Male or Female. School related factors Academic E ngagement (question 18) Academic engagement was measured through the School Success Profile domain of s chool e ngagement (Bowen & Richman, 2005) This is a 3 item student self report subscale which measures whether students find school fun and exciting, look forward to learning new things at school, and look forward to going to school (Powers et al., 2005) Responses range d from (1) not like me to (3) a lot like me for a total score from 3 and 9 The questions are worded so that higher numbers indicate a promotive effect, while lo wer numbers indicate risk, and the scale has been shown to have a good internal consistency ( = .80 ). Within this dataset, internal consistency was found to be good as well ( = .79). Academic E xpectations (questions 6, 7, and 16) Academic expectations was measured from a single item that is derived from the work of Abu Hilal (2000) which asked how far students expected to go in school, with

PAGE 65

55 responses ranging from (1) not finish school to (8) complete an advanced degree Similar measures of expectations have yielded a strong relationship with academic achievement, correlation coefficients above .60 (Sanders, Field, & Diego, 2001; Tavani & Losh, 2003) This question was also combined with the work of Dandy and Nettlebeck (2002) who asked student which grades th ey would be satisfied receiving and if students thought they would finish high school. Academic S elf E fficacy (question 17) Self efficacy was captured through the academic subscale from the Self Effica cy Questionnaire for Children ( SEQ C; Muris, 2001), which measures students feelings about their ability to be successful in school and display appropria te learning behaviors. Student s were asked to rate their competence on each question using a 5 item Likert scale (1 not at all to 5 very well ), with responses summed for a total score. The version that was used is a 7 item scale that was modified slightly from the original version b y Su ldo and Shaffer (2007) This version was adjusted to account for American speech and subsequently administered to American youths. The modified version was found to retain good internal consistency ( .8 2) which was also found within the current data set to be good ( .81 ) Attendance (school records) was obtained through school records. School absences are recorded for each period of the day. To acquire the average number of absences per semester by student, the total number of class absences were subtracted from the number of total classes per semester and then divided by the number of total classes per semester. For a school with a six period day, the formula is:

PAGE 66

56 1080 A 1080 For a four period day, the formula is 720 A 720 Educational S upport (questions 19 and 20) Educational support was measured through the School Success Profile domain of parent educational support and home academic environment (Bowen & Richman, 2005) Home parent educational support is a 6 item scale which captures whether adults in the home encourage/support school and work activities, he lp get needed supplies, and offer help with homework or special assignments. Home academic environment is an 8 item scale capturing whether students discuss their courses or programs at school, their school related activities, current events and politics, and their plans for the future with the adults who live in their home. The responses ranged from (1) never to (3) more than twice. The questions are worded so that higher numbers indicate a promotive effect, while lower numbers indicate risk The parent educational support scale has been shown to have a while the home educational environment has an internal consistency of .87. Using the current data, both the parent educational support ) and home educationa When the scales were combined in this study A further analysis showed that the internal reliability would not increase if any of the items were deleted.

PAGE 67

57 Grade R etention (question 5) Grade retention was measured through a single item as used in the School Success Profile (Bowen & Richman, 2005) which asks how many times they have been retained in school. In previous research this factor has been dichotomized as 0 for no retentions and 1 for any retentions (Woolley & Bowen, 2007) Home work (question 2) Homework used a question developed to measure the proportion of homework completed (Cooper, Lindsay, Nye, & Greathouse, 1998) The 6 responses range (1) none to (6) all This single item has been shown to be moderately correlated with achievement ( r = .31) and is a stronger predictor of grades than time spent studying or comple ting homework. Music Instruction (questions 8 and 9) Music involvement was captured by using questions developed by previous researchers. Schellenberg (2006) asked how many years a student regularly played music (with or without lessons). When asked this w ay, this mea sure has a small effect size when looking at s chool grades ( r = .22). Fitzpatrick (2006) also looked at music, but classified students as instrument playing (1) or non instrument playing (0). If students were ever in volved in band, jazz, ch orus, or other school music classes they were classified as music playing, otherwise they were coded as non instrument playing. This measure has been shown to be related to achievement in math, reading and science School Behavior (school records and question 23) School behavior was measured through the School Success Profile domain of trouble avoidance (Bowen & Richman, 2005) This 11 item scale looks at whether

PAGE 68

58 students have avoided problem behaviors in school over the past 30 days. The responses range from (1) never to (3) more t han twice with a total score ranges from 11 to 33. The questions are worded so that higher numbers indicate a promotive effect, while lower numbers indicate risk. The scale has been shown to have a s trong internal consistency in previous research ( 82) and within the current data ( 81). School Belonging (question 16) School belonging was measured through the School Success Profile domain of school satisfaction (Bowen & Richman, 2005) This 4 item scale looks at whether students enjoy school, get along with teachers and peers, and feel they are gett ing a good education. Responses range from (1) not like me to (3) a lot like me with a total score ranges from 4 to 12. The questions are worded so that higher numbers indicate a promotive effect, while lower numbers indicate risk. The scale has been shown to have a good internal consistency in previous literature ( ) and w ithin the dataset was calculated at .66. School M obility (question 5) School mobility was asked as the number of times a student has changed schools in the last three years, excluding moving from middle to high school which is a regularly scheduled trans ition (Dunn et al., 2003) School R elevance (question 4) Alpert and Dunham (1986) weakly associated so it was dropped from their analysis. This was asked as a dichotomous yes or no question.

PAGE 69

59 School S afety (question 21) School safety was measured through the School Success Profile domain of school safety (Bowen & Richman, 2005) This 11 item scale captures whether students attend a school with a low crime level, few problem behaviors, and few bullying behaviors. The responses ranged from (1) not a problem to (3) a big problem with a subscale total s core ranges from 11 to 33. T he questions are worded so that higher numbers indicat e a promotive effect, while lower numbers indicate risk. The scale has been shown to hav e a strong internal consistency ( = .88). A comparable internal consistency was also found within the current data ( = .87). Supportive School Environment (question 24) To measure whether students feel they have supportive school environment the School Success Profile domain of learning climate was used (Bowen & Richman, 2005) This is a 7 item subscale which measures whether come first, adults at school affirm and care about them, and every student is valued Responses range from (1) strongly disagree to (4) strongly agre e for a total score between 7 and 28 The questions are worded so that higher numbers indicate a promoti ve effect, while lower numbers indicate risk. The scale has been shown to have a good measuring the internal consistency with the current data showed this to be strong as well Teacher S upport/ R elationship (question 22) To measure teacher support and relationships the School Success Profile domain of teacher support was used (Bowen & Richman, 2005) This is an 8 item subscale which

PAGE 70

60 measures whether students perceive teachers at their school as supportive, as caring about them and their academic success, and as expecting them to do their best. These areas have been combined because the scale being used captures items of both teacher support and relationships, which has been acknowledged by Bowen and Richman (2005) in their analysis of the validity and reliab ility of the School Success Profile. Responses range from (1) strongly disagree to (4) strongly agre e for a total score between 8 and 32. The questions are worded so that higher numbers indicate a promotive effect, while lower numbers indicate risk. Th e scale has been shown to have a stro = .89). Instrument Pilot Testing Each set of questions addressing the individual school related factors were compiled into a single instrument. During the pilot testing phase, the instrument was examined by 19 high school age students for readability, structure, completion time, and overall assessment of the instrument The students ranged from 9 th 12 th grade and were aged 15 19 (M = 16.01, SD = 1.2). The language ability ranged from complete fluency in English (n = 15) to various degrees of English speaking ability (n = 4). In addition, 17 (89.5%) students were female and 2 (10.5%) students were male, with the students ha ving a wide range of grade point averages from 0.50 to 4.00. A majority of the respondents were Hispanic (63.2%), followed by White (15.8%), Black/African American (10.5%), Multi Racial (5.3%) and American Indian/Alaskan Native (5.3%). The time to comple te the instrument ranged from 5 minutes to 24 minutes, with a mean time of 11.79 minutes (SD = 4.9) None of the respondents reported having any difficulty with understanding the questions or answers. However, 5 respondents

PAGE 71

61 option for question 8 which was incorporated into the final instrument. Overview of Risk and Protective Scores Risk/Protective Factor Scores Each of the factors are scored by giving a ri sk score (1), a protective score ( 1), or a non risk score (0) depe nding on the The risk factors will be added to create a summated score. For the cumulative risk model only risk scores will be added, while for the additive risk model both risk scores and protective sco res will be added together. Risk and protective facto rs selected will be converted to risk or protective scores and coded as 1 for present and 0 for absent based on their relationship to achievement For those factors which are continuous in nature, th e samples will be divided into risk, non risk, and promotive. This type of division was done in previous stud ies on continuous factors th r ough defining risk status by dividing the sample as risk for the lowest 25% of the sample, promotive for the top 25% of the sample and medium or non risk for those in the respondents in the middle 50% (Bowen, 2006; Kinar d, 2001; Sameroff et al., 2003) Some scores only have a promotive effect, while others show solely a risk effect. For these scores, they are coded depending on the nature of the relationship with cumulative GPA. In addition, s ome risk scores are curvilinear in nature and in order to be included in the analysis there they will be divided into individual risk scores that can be assigned risk or protective status Cumulative Risk Model Summing individual risk factor scores to create a cumulative risk model has been used by numerous researchers, with the computation of risk scores and analysis being

PAGE 72

62 simi lar in each case (Conners, Bradley, Mansell, Liu, Roberts, Burgdorf, & Herrel, 2003; Corpaci, 2008; Gassman Pines & Yoshikaw a, 2006; Gutman et al., 2003; Gutman et al., 2002; Sameroff, 1985) Once a composite score is created, they are regressed against the achievement related o utcome to determine the significance of the overall model. Additive Risk Model O thers have created an additive risk model by using the same methodology that was applied previously to cumulative risk model creation (Prelow & Loukas, 2003) The difference lies in including both risk and protective factors in the model The individual factors were looked at in their relationship to school problem behaviors and academic achievement. It was found the more risk factors present, the lower the academic achievement and the higher number of problem behaviors at school. However, the relative risk was reduced as more protective factors were present, implying that specific protective factors could o ffset the effects of risk. The final model includes scoring protective factors as a 1 and risk factors as a +1, creating an overall score which accounts for the effect of each individual factor. Data Collection Procedure Approval for data collection incl uded both an application for the University of South Florida Institutional Review Board (IRB) and the Pinellas County Schools Research and Accountability department (see Appendix E) The next step required contacting the three schools and acquiring permis sion from each principal to conduct research on their school site. Once these permissions were granted, a schedule for recruitment and survey administration was arranged with each school. In order to recruit students, a letter was sent home with each stu dent at the school (See Appendix D ). This

PAGE 73

63 letter included a description of the study and a consent form for participation. The letter and giving permission to combin e these results with school records. Specifically, the school records included grade point average, number of absences, race, grade in school, free/reduced lunch status, FCAT scores, and the number of behavioral referrals. Parents were given a two week p Those students whose parents returned a consent form were contacted and offered participation in the study. In addition to parental consent, child assent was also required of students for their involvement in the study. Child assent was obtained at the same time as survey administration, in order for the researcher to answer any specific questions that may arise regarding the study. Administration of the surveys occurred over a one week period at all three schools. The surveys were administered in groups of 5 20, during with school records to form a complete data set.

PAGE 74

64 Chapter Four Statistical Analy sis Prior to analysis the raw data was examined to look for missing values, distribution of data, and accuracy of data entry. A frequency analysis was run on each variable in the dataset and found no missing data in any of the variables that were to be included in the analysis. This analysis was also used in conjunction with descriptive data to verify that all data fell w ithin the range of the responses, and no outlying data was found. In addition, the accuracy of data entry was checked. Ten percent (n = 22) of the sample was randomly selected and given to a third party to verify the accuracy of the data that had been entered. Each survey contains 87 items, which results in 1,914 total entries being checked. Four errors in data entry were found dur ing the check, resulting in an accuracy of 99.79%. These data entry errors were corrected in the data set. Analytic Approach This section will discuss the analytic approach that was taken with each of the five questions that were posed earlier. In order Which school factors impact academic achieve ment among high school students oes the cumulative risk model predict academic achieve was analyzed using a two part process. Initially, an additive risk index was creating by summing the risk/promotive scores. These scores were then regressed against cumulative GPA using both standard linear and log oes the additive risk model predict academic achievement among high school stud creating a cumulative risk index which was the summed total of risk scores. The

PAGE 75

65 resulting CRI was then regressed again st cumulative GPA using both standard linear and s the cumulative risk model or additive model a better predictor of achievement levels comparing the regression results that were used to answer questions two and three. hat is the optimal number of academic domain risk factors for distinguishing between students who are at risk and not at risk ROC curve analysis which looks at the interplay between sensitivity and specificity. Descriptive Analysis For each factor that was examined in the study, a descriptive analysis was run, which included mean, range and standard deviation. Academic engagement had a mean score of 6.17 ( SD = 1.62), and a range from 3 9. Overall, attendance had a mean rate of 88.90 (SD = 11.75), with the lowest reported attendance at 39.88% and the highest rate of attendance reported at 99.42%. School behaviors was found to have a mean score of 15.59 (SD = 3.77), and ranged from 11 30. The mean score for school mobility was 1.59 (SD = .97) and ranged from 1 4, while grades repeated had a mean of 1.18 (SD = .46) and also ranged from 1 4. Ninety seven percent of students found that high school would get t hem the job they wanted (SD = .18), and the needs scale had a mean score of .51 (SD = 1.06) and scores ranged from 0 7. Finally, a cademic self efficacy was found to have a mean score of 26.46 (SD = 5.08) and had a range from 7 35. Table 6, in Appendix A, sh ows more detailed information for each of the factors that were included in the study design Initially, demographic and control variables were examined to determine the relationship between the sample and the population from which it was drawn. For t his

PAGE 76

66 analysis, the data was split into the sample and the population minus the sample This was done to ensure that data from the sample group was independent from the population for the comparison. The descriptive information for each group is displayed in Table 7 in Appendix A. Findings indicate the sample and the population were not significantly different on free/reduced lunch status ( 2 (1, N = 2212) = .31 p = .579) and grade in school ( 2 (3, N = 2212) = 7.593, p = .055). However, the sample when compared to the entire population of the three schools did differ on cumulative GPA, t (274.1) = 3.37 p < .001, with the sample being slightly higher (M = 2.85, SD = .77) th a n the population (M = 2.67, SD = .85). There were also differences found with the gender of the sample having a higher percentage of female respondents (79.7%) than the overall group ( 48 .0%), 2 (1, N = 2 21 2) = 78.73 p < .001. Finally, race was also shown to be different between the sample and the overall population, 2 (5, N = 22 12) = 32.32 p < .001. The sample had a higher percentage of Hispanic students (23.5%) than the population (11.0%) and a lower percentage of African American students (9.2%) when compared to the overall population (16.1%). Relationships a mong Factors an d Cumulative GPA Scatter plots were run individually on each factor, with the factor on X axis and cum ulative GPA on the other Y axis. Examining th ese outputs revealed all relationships w ere linear suggesting that a correlation analysis could be perform ed. Each of the factors was inspected to determine the correlation between each factor and cumulative GPA as well as the relationship between factors As seen in Table 8 in Appendix A s ignificant relationships ( p < .01) were found between 11 factors and cumulative GPA. The strongest relationships were found with academic expectations ( r = .606), attendance

PAGE 77

67 ( r = .533), academic self efficacy ( r = .463), grade retention ( r = .399), and SES ( r = 375). Academi c expectations, attending more frequently, and academic self efficacy all had positive relationships with cumulative GPA, suggesting that as each factor score got higher, so did GPA. However, the more times students had been retained a grade and the highe r poverty level of a student, the more likely they were to have a lower cumulative GPA. Medium to strong negative correlations were found with achievement and both school behaviors ( r = .364) and school mobility ( r = .353). This meant the more problem s chool behaviors that a student d isplayed and the more times a student moved, the lower their GPA Additionally, playing music playing ( r = .281), having more educational support ( r = .281), and doing a higher percentage homework ( r = .280) all were equated with increased cumulative GPAs. Finally, school belonging ( r = .157) showed a small to medium, but positive effect on achievement. This suggests that students who felt more closely connected to their school did better with schoo l GPA than those who were not connected. When looking at the factors relationship to each other, none of the correlations was above .70 However, there were several strong relationships worth noting, with the strongest relationships found between academi c self efficacy and academic expectations ( r = .419), school behaviors and academic self efficacy ( r = .413), and between academic expectations and grade retentions ( r = .412). Other strong relationships existed between academic self efficacy and homewo rk ( r = .398), academic expectations and attendance ( r = .389), and between school belonging and academic self efficacy ( r = .363). On the opposite end, it was interesting to note there was not a significant

PAGE 78

68 relationship with school behaviors and grade ret ention ( r = .013). Factors that were found to be not significantly related to cumulative GPA were not added to the cumulative risk or the additive risk models. These factors include academic engagement ( r (217) = .051, p = .452), school relevance ( t (215) = .433, p = .438), school safety ( r (217) = .009, p = .890), supportive school environment ( r (217) = .027, p = .690), and teacher support ( r (217) = .124, p = .069). The four demographic factors that were shown in previous research to impact achievement were also tested to determine if there was a need to control for these variables in the final analysis. Gender ( r pbi = 121 p = .075) and living situation ( F (3,213) = 1.907, p = .129) were shown to be non significantly related to cumulative GPA However, both SES and race were shown to have a significant relationship with cumulative GPA. When looking at t he two measures of SES both the needs scale ( r (217) = .294, p < .001) and free/reduced lunch ( r pbi = .310 p < .001) were found to be stati stically significant w hen compared to cumulative GPA. A significant relationship was also found when looking at race ( F (3,213) = 9.385, p < .001) with Asian students having the highest GPA (M = 3.43, SD = .43), followed by White (M = 3.03, SD = .72), His panic (M = 2.69, SD = .79), Multi Racial (M = 2.42, SD = .54), Black (M = 2.17, SD =.66) and American Indian (M = 2.05, SD = 0.00). Question 1 Individual Risk and Protective Factors Which school related factors impact academic achievement among high school students? Each variable was examined to determine if they met the criterion for a risk or promotive factor. This was done based on theoretical perspectives, previous research, or empirically determined. To determine which factors should be include d in the

PAGE 79

69 cumulative and additive risk models, f actors must meet three criteria. First e ach factor that is included in the models m ust be correlated with GPA For the first criterion, c orrelation matrices were run on each factor computing its relat ionshi p to grade point average. The second criterion involves showing a significant difference in the outcomes between adolescents in the present versus absent risk group This step is determined by performing a t test or a one way ANOVA on each factor between the risk and non risk groups (for the CRI) and the risk, non risk, and promotive groups to establish if there is a significant difference in the outcome measure of cumulative GPA for these groups, so that those families who experienced the risk factor pe rform significantly differently than families wi thout that environmental risk (Sameroff, 2003) Finally, a linear regression analysis was run and those factors that made a unique contribution to the overall model were included in the respective index. Continuous variables were classified as risk for the bottom 25% 30% of scores, the middle 50% were considered non risk (0), and the highest 25% were coded as promotive (Sameroff et al., 2003) As a result, each factor was coded into dichotomous variables for inclusion in the CRI (1 = presence of risk factor, 0 = absence of risk factor) or three categories for inclusio n in the ARI ( 1 = presence of risk factor, 0 = non risk, and +1 = presence of promotive factor). In the following section each variable will be discussed in how the risk or promotive scores were assigned for both the Additive Risk Index (ARI) and the Cumu lative Risk Index (CRI) (Bowen & Richman, 2005; Sameroff, 1985, 2003)

PAGE 80

70 Data Aggregation Socioeconomic Status (SES) SES was measured on the survey instrument using two distinct items, Running a t test revealed a significant relationship between free/reduced lunch and GPA. There was a higher GPA for those that did not receive free lunch than those that did, t (215) = 4.782, p < .001. A separate analysis was run on the needs scale and found that a negative relationship with cumulative GPA, r (217) = .294, p < .001 and a positive relat ionship with free/reduced lunch, r(217) = .265, p < .001. These two items were combined in order to account for the problem s of validity that are cautioned b y researchers when using a single measure response item (Bergkvist & Rossiter, 2007; DeVellis, 2003; Loo, 2002) but at the same time enabled the inclusion of free/reduced lunch which is often used as a proxy measure for poverty. Combining the dichotomous free/reduced lunch and the summed needs scale was accomplished by converting each measure to a standardized score, adding them together, and then taking the average of the two scores. This gave equal weight to each mea sure, and the final SES score was shown to have a strong relationship with both the needs scale ( r (217) = .805, p < .001) and free/reduced lunch status ( r( 217) = .763, p < .001). In addition, there was also a significant negative relationship with cumulat ive GPA, r(217) = .375, p < .001. Academic Expectations Each of the three individual measures of academic expectations showed a significant relationship with academic achievement. The 8

PAGE 81

71 r (217) = .531, p < 001), the 5 (r (217) = .390, p < .001), and the 5 r (217) = .419, p < .001) were combined to created a single academic expectations measure whic h was shown to be significantly related to cumulative GPA ( r (217) = .606. p < .001). Each of these questions has been used in previous research as a measure of academic expectations (Abu Hilal, 2000; Dandy & Nettelbeck, 2002) Scores on the overall measure ranged from 6 17, with a mean score of 14.47 (SD = 2.49). Responses were then aggregated according to the ARI with 6 14 (33.6%) as risk, scores 15 16 as non risk (49.3%), and 17 (17.1%) as promotive. For the CRI, scores were computed from 6 14 as risk (33.6%) and 15 17 as non risk (66.4%). An ANOVA was run and found a significant difference among risk scores and the relationship between cumulative GPA, F (2, 214) = 35.861, p < .001. A further examination using a Bonfe rro risk, non risk, and promotive scores, indication each level has a unique relationship with academic achievement. Academic Self Efficacy Academic self efficacy was shown to be related to cumulative GPA, r(217) = .463, p < .001. This measure ranged from 7 35, with a mean score of 26.46 (SD = 5.08). Scores for the ARI were divided with 7 23 (29.0%) as risk, 24 30 (48.4%) as neutral, and 31 35 (22.6%) as promotive. The CRI scores were divided as 7 23 (29.0%) as risk, and the rest as non risk. Further analysis using an ANOVA showed this factor met the second criterion for use as a risk factor, with a significant difference among the groups in the relationship to achievement, F (2, 214) = 24.875, p < .001.

PAGE 82

72 Attendance Rate Attendance rate was also tested to determine if it met the criterion for inclusion in the ARI and CRI models. Initially, a correlation analysis revealed a significant relationship to cumulative GPA, r (217) = .533, p < .001. Accor ding to previous research, this factor was coded as risk for those who attended less than 80% of the time, and non risk for attendance above 80% (Crean et al., 2001; Klem & Connell, 2004) A t test showed that attendance rate met the second criterion as those with the risk factor had a lower GPA ( M = 1.97, SD = .725) than those who attended more than 80% of the time ( M = 3.01, SD = .672), t (215) = 7.962, p < .001. Educational Support Parent educational support ( r (217) = .268, p < .001) and home academi c environment ( r (217) = .239, p < .001) covered topics that were similar. An initial analysis revealed that the items were highly correlated ( r (217) = .702, p < .001), suggesting each variable was measuring the same factor. As a result these two measures were combined into a single measure of educational support, which was also related to cumulative GPA, r (217) = .289, p < .001. The final scores ranged from 14 42, with a mean score of 31.04 (SD = 6.47). For inclusion in the ARI, the bottom 25% of scores were classified as risk (14 26, 24.9%), the middle range was grouped as non risk (27 36, 53.4%), and the upper range of 37 42 (21.7%) were coded as promotive. Running an ANOVA revealed a difference between risk, non risk, and promotive scores, F (2, 214) = 10.547, p < .001. For the CRI analysis, scores 14 26 (24.9%) were listed as risk and 27 42 as non risk (75.1%).

PAGE 83

73 Hom ework Percentage of homework completed was shown to have a significant relationship with cumulative GPA, r (217) = .281, p < .001. To examine where the differences were, an ANOVA revealed that there was a difference between none and all ( p = .003), F (4, 21 2) = 4.756, p < .001. The scores were classified as the recoded item revealed a difference between risk and both non risk and promotive, F (2, 214) = 6.893, p = .001. However, there was no difference between non risk and promotive, suggesting that this factor is only a risk factor and has no promotive value that is unique when compared to non risk. The new rating was then followed up using a t test, revealing a difference between risk and non risk factor in regards to cumulative GPA, t (215) = 3.117, p < .001. The final coding for both the ARI ( 1) and the CRI (1) risk (0). Grade Retention T he number of grades a student repeated was related to cumulative GPA, r(217) = .399, p < .001. An ANOVA confirmed previous research which classified any grades retained as risk, F (2, 214) = 20.828, p > .001 (Woolley & Bowen, 2007) Significant differences were found between no retention s and one ( p < .001), two ( p < .001), and three or more ( p p = .368). As a result, students who were retained at all were classified as risk (ARI = 1, CRI = 1) and students not having been retained were coded as non risk (n = 184).

PAGE 84

74 Music Playing Two items were used to determine if students were music playing. B oth the types of in school music classes ( r (217) = .183, p = .004) and years playing music ( r (217) = .197, p = .007) were related to final cumulative GPA. Students were classified as music playing if they either took in school courses or had played music regularly. A t test revealed that music playing students displayed a higher GPA than non music playing students, t (215) = 4.361, p < .001. This confirms that music playing was a promotive only factor, and was added to the ARI model only. School Behavior s School behaviors were measured using both number of referrals and the school behaviors subscale from the instrument. Both number of referrals ( r = .387, p < .001) and school behaviors ( r = .364, p < .001) were significantly correlated with grade point average. These two measures were combined into one single measure of school behaviors. Each score was converted to a z score and averaged for a final score. The final measure of s chool behaviors displayed a significant relationship to GPA r (217) = .4 44 p > .001. The mean score for this measure was 0.00 (SD = .86 ), with a range of .87 to 4.14 . The responses were coded as .32 to 4.14 ( 24.0 %) for risk .59 to .31 as non risk ( 50.7 %), and .87 to .60 (25.3 %) as promotive. Running an ANOVA revea led a significant difference among risk, non risk, and promotive, F (2, 214) = 30.700 p < .001. risk and promotive were not significant ( p = 853 ), though there was a significant difference between both non risk/promotive and risk ( p < .001). This suggest school behaviors should be viewed as risk only, which was confirmed using a t test which revealed a lower GPA for the risk

PAGE 85

75 group over the non risk group, t (215) = 7.759 p < .001. The fin al coding is then risk (ARI = 1, CRI = 1) and non risk ( CRI and ARI = 0). School Belonging The school belonging scale was revealed to have a significant relationship with r (217) = .157, p = .021. This measure had a range from 5 12, with a mean score of 9.75, (SD = 1.66). The data were coded as 5 8 for risk (25.8%), 9 11 as non risk (57.1%), and 12 as promotive (17.1%). Further analysis showed this factor to have a promotive value only using an ANOVA ( F (2,214) = 4.753, p = .01 0, which revealed a difference between promotive and risk ( p = .018), as well as promotive and non risk ( p = .014). However, no difference was found between risk and non risk ( p = 1.00). A t test confirmed the difference between GPAs for those who were c lassified as promotive and those who were not, t (215) = 3.072, p = .002. For the ARI, promotive was coded as +1, and everything else was coded as 0. This factor was not included in the CRI. School Mobility School mobility was found to display a signific ant relationship with cumulative GPA, r (217) = .353, p < .001. An one way ANOVA was then used to take a further look at the breakdown of this variable and showed a relationship between no moves and two moves ( p = .010) and three or more moves ( p < .001), F(3, 213) = 10.193, p < .001. As a result of this analysis, a risk score was compiled for any students who moved two or more times (ARI = 1, CRI = 1), with everyone else being coded as non risk. A t test was then run, which showed that students with a risk status had a lower GPA ( M = 2.31) than those who had no moves or one move ( M = 2.96), t (215) = 4.914, p < .001.

PAGE 86

76 Controlled Factors Race, Gender, and SES were used as control variables in the final model. Dummy variables were created for Race in order to include this factor in the linear regression analysis. Gender was included to control for the difference between the proportion of males and females in the sample and the population. In addition, SES was examined and found to have a skewness (1.52) and kurtosis (3.14) values above the desired range. In order for this factor to be included as a control variable, the scores were conve rted to T scores to remove any negative numbers, and then a log transformation was performed. This resulted in skewness (.98) and kurtosis (.52) values within acceptable ranges. Unique Contribution of Factors The final condition for inclusion in the cu mulative and additive risk indices is that the factor must also contribute uniquely to the overall model. Determining this was accomplished by putting all of the risk and promotive factors that met the first and second criterion into a linear regression m odel. For the CRI, of the nine risk factors, five significantly contributed to the model above a nd beyond the control variables, F ( 12 20 4 ) = 80. 14 p < .001. Academic engagement ( p = .616), educational support ( p = 678), percentage of homework completed ( p = .528), and school mobility ( p = .814) were not significant in this model Removing these factors from the regression model made no difference in the R 2 or significance of the model, so they were deleted This left percentage of day s present, academic self efficacy, academic expectations, grades repeated, and school behaviors. The ARI was nearly identical to the CRI in that it was significant, F (13, 203) = 80.99, p < .001, and contained the same factors with the addition

PAGE 87

77 of music playing being a significant contributing promotive factor to cumulative GPA. Table 9 in Appendix A gives a more detailed examination of the overall regression models for the CRI factors and Table 10 provides details for the model including the ARI factors. Cumulative G PA The distribution of the cumulative GPA was also examined prior to analysis. The overall sample had a mean GPA of 2.85 (SD = .77) and ranged from 0 4. Additional analysis revealed the distribution of the GPAs was within ranges that would be conside red acceptable for both skewness and kurtosis. The skewness was .74 and the kurtosis was .023. The distribution of the data also showed that three students had a cumulative GPA less than .8, making them moderate outliers. A further analysis revealed t hat 94.9% of the sample fell within 2 standard deviations from the mean, 1.4% were above 2.56 standard deviations from the mean, and no scores were above 3 standard deviations from the mean. This information is summarized in Table 11 (see Appendix A). Computation of the Cumulative Risk Index The technique applied for computing the CRI involved summing the five factors that were found to contain a unique risk component. These were percent of present, academic expectations, academic self efficacy, grade retention, and s chool behavior As a result, an individual risk score was calculated for each student. The mean CRI score was 1. 17 (SD = 1.28 ), with scores ranging from 0 to 5 Forty two percent of the students had a risk score of 0 (n = 92) fol lowed by 23% (n = 50) with a score of 1, 16.6% (n = 36) had a score of 2, 13.4% (n = 29) had a score of 3, 2.8% had a score of 4

PAGE 88

78 (n = 6), and 1.8 % (n = 4) had a risk score of 5 A further analysis of the CRI scores showed a positively skewed distribution but values of both skewness (.92) and kurtosis (.05 ) were within acceptable ranges. In addition, an examination of the scores using box plots showed no scores were considered outlier s Outliers were also examined though the distribution of the scores, which found that 96.8% of the scores were within 2.00 standard deviations from the mean, and no scores were found over 2.58 standard deviations from the mean. A reliability analysis was run on the CRI, which produced a of .6 5 The d eletion of any items would add no addition al value. A summary of descriptive information for the CRI is available in Table 11, Appendix A Computation of the Additive Risk Index The computation of the ARI was done by totaling the values of the risk, non r isk, and promotive scores and giving each student an individual additive score. The six factors that which had a direct and unique contribution to cumulative GPA were included in this calculation These were percent of days present, music playing, academ ic expectations, academic self efficacy, grade retention, and school behaviors. The final scores range from 5 to + 4 with a mean score of 06 (SD = 1.95 ). E xamining the normality of the ARI the distri bution found the skewness at 1 6 and the kurtosis a t .50 which are both within normal ranges. Box plots showed two values were considered moderate outliers ( 5) which was confirmed by examining the distribution of the scores. Finally, the reliability of the ARI was calculated at .6 9 using Ch r alpha. The descriptive information is summarized in Table 11 (see Appendix A).

PAGE 89

79 Descriptive Analysis of the CR I and ARI Factors by Risk Group For each factor that was included in the final Additive Risk Index (ARI) and Cumulative Risk Index (CRI) a descriptive analysis was performed (mean, standard deviation, and range). While the average cumulative GPA for this sample was 2.85 (SD = .77), the risk, non risk, and promotive groups differed among each factor. For example, the range of GPAs for academic self efficacy went from 2.34 (SD = .81) for the risk group, to 3.02 (SD = .62) for the non risk group, and finally 3.28 (SD = .50) for the promoti ve group. A full analysis of the data prior to and after aggregation is presented in Appendix A in Tables 12 and 13 Table 12 displays this data for the Additive Risk Index, while Table 13 provides details for the Cumulative Risk Index. Question 2 Cu mulative Risk Model Does the cumulative risk model predict academic achievement among high school students? In order to test the predictive ability of the CRI, a standard linear regression was run with cumulative GPA as the dependent variable and the CRI as the independent variable. The impact of race, gender, and SES were controlle d in the analysis as well by entering these factors into the first step of the regression analysis, followed by the CRI in step two. Analysis of assumptions was done using bot h before and after regression by looking at the data, outputs, and residuals. As previously reported, a log transformation was used to reduce the skewness and kurtosis of SES. All other factors were found to be within normal ranges a nd were entered untran sformed, and race w as entered using dummy coding. Multicolinearity was examined using both correlations and VIF tolerances. The highest correlation was

PAGE 90

80 between Hispanic and SES ( r = .409, p < .001), followed by CRI and SES ( r = .405, p < .00) which were low enough to raise no concerns regarding multicolinearity. An examination of the VIF tolerances showed a range from 1.01 to 1. 56 Outliers were which revealed there were no outliers. A Durbin Wa tson value of 1. 923 shows no discernable pattern of scores into clusters, suggesting the errors are independent of each other Examination of the residuals showed the data met assumptions of linearity, h omoscedasticity and distribution. Figures 4 throug h 6 in Appendi x A show the distribution of the standardized residuals None of the 217 cases had missing data. Table 1 4 shows the corr elations between the variables, while Table 15 displays the unstandardized coefficients ( B ) and intercept, the standardi zed regression coefficients ( ), the R 2 the adjusted R 2 and VIF tolerances (See Appendix A) R for regression was found to be significant, F(8, 208) = 72.99 p < .001, with the R 2 value at 566 In the first step of the analysis, the linear combination of the three control variables race, gender, and SES produced an R 2 value of .229, with an adjusted R 2 of .203. When adding the CRI to the regression analysis, the R 2 was raised to .5 56 and adjusted R 2 rose to .5 50 resulting in a change in R 2 of .3 37 and a change in adjusted R 2 of .3 47 This suggests that above and beyond the influence of race, gender, and SES, cumulative risk accounted for an additional 3 4 % (3 5 % adjusted) of the variability in cumulative GPA. The size and direction of the r cumulative GPA goes down. Figure 7 illustrates this in graphing the number of risk factors and cumulative GPA Although significant in step one of the model Gender is no longer signi ficant in the final CRI model. The relationship between Black, Asian Multi

PAGE 91

81 racial, SES, and the CRI appears to mediate the relationship b etween cumulative GPA and gender. Question 3 Additive Risk Model Does the additive risk model predict academic a chievement among high school students? Testing of the ARI involved running a standard linear regression with cumulative GPA as the dependent variable and the ARI as the independent variable. The impact of race, gender, and SES were controlled in the analy sis as well by entering these factors into the first step of the regression analysis, followed by the ARI in step two. Analysis of assumptions was done using both before and after regression by looking at the data, outputs, and residuals. Prior to running the regression, a log transformation was used to reduce the skewness and kurtosis of SES. All other factors were found to be within normal ranges and were entered untransformed. R ace was entered using dummy coding. Multicolinearity was examined using both correlations and VIF tolerances. The highest correlation was between the Hispanic and SES ( r = 409 p < .001), followed by ARI and SES ( r = .40 8 p < .001). These correlations were well below any values which w ould indicate multicolinearity, indicating this assumption was met. An examination of the VIF tolerances showed a range from 1.01 to 1. 56 distance, which revealed no outliers. A Durbin Watson value of 1. 76 shows no pa ttern of scores clustering, suggesting the errors are independent of each other. Examination of the residuals showed the data met assumptions of linearity, h omoscedasticity and distribution. Figures 8 through 10 in Appendix A show the distribution of th e standardized residuals. None of the 217 cases had missing data.

PAGE 92

82 Table 16 shows the correlations between the variables, while Table 17 displays, the unstandardized coefficients ( B ) and intercept, the standardized regression coefficients ( ), the R 2 the adjusted R 2 and VIF tolerances (See Appendix A). R for regression was found t o be significant, F(8, 208) = 33.43 p < .001, with the R 2 value at 563 In the first step of the analysis, the linear combination of the three control variables race, gender, and SES produced an R 2 value of .229, with an adjusted R 2 of .203. When adding the ARI to the regression analysis, the R 2 was raised to 563 and adjusted R 2 rose to 546 resulting in a change in R of .3 33 and a change in adjusted R 2 of .3 4 3 This suggests that above and beyond the influence of race, gender, and SES, cumulative risk accounted for an additional 3 3 % (3 4 % adjusted) of the variability in cumulative GPA. The size and direction of the re lationship suggests that as the additive score cumulative GPA goes up, and conversely as the risk scores increase then GPA decreases Figure 11 illustrates this in graphing the additive total of risk and promotive factors and cumulative GPA. While a significant bivariate correlation was found between cumulative GPA and was found between cumulative GPA and gender ( r = .12, F (8, 209) = .13 p = .0 4 ), Black ( r = .282, F ( 8 2 09 ) = 2. 25 p > .001), Hispanic ( r = .122, F (8, 209) = .39 p = .04 ), and SES ( r = .390, F (8, 209 ) = 4.66 p < .001), using a post hoc correction only Black and Asian were significant in the final regression model. The relationship between Black, Asian and the ARI appears to mediate the relationship between cumulative GPA and gender Hispanic, and SES.

PAGE 93

83 Question 4 Cumulative R isk M odel versus A dditive M odel Is the cumulative risk model or additive model a better predictor of achievement levels for high school st udents? Two methods were employed i n order to examine if the CRI or ARI were better predictors of achievement levels. A logistical regression was run to examine which model correctly classified students according to their risk status and the linear regres sions previously run were also examined to determine which model accounted for more variability in cumulative GPA. In order to use a standard logistical regression, the dependent variable must be discrete, so cumulative GPA was divided as risk for student s whose GPA was less than a 2.00 GPA and non risk for GPAs above 2.00 This cut off point was selected because in order to graduate from high school in Florida a student must have a GPA above 2.0 (Florida Senate, 2007) Cross Validation The regression models for both the CRI and the ARI were cross validated by splitting the sample approximately in half (n = 115) and running a linear regression the selected portion of the sample. The resulting regression equation was used to compute a predicted score for the other half of the sample (n = 102), which was then compared to their actual scores. This procedure w as completed for both the CRI and ARI regression models. Results of this cross validation indicate a high correlation between the predicted score for both the CRI ( r (102) = .743, p < .001) and the ARI ( r (102) = .733, p < .001). This suggests the factors that are included in the model are appropriate and there is minimum shrinkage.

PAGE 94

84 Model Comparison Examining the variability calculated using linear regression revealed t he model which included the CRI as a predictor accounted for 1% more variability than the ARI model. A second check of running two logistical regression models using the CRI and ARI as predictors, while controlling for race, gender, and SES was also done. The CRI model was found to be significantly different than t he original model with no predictors ( 2 (8, N = 217) = 76.15 p < .001) and adding controlled factors alone ( 2 (1, N = 217) = 48.71 p < .001). There was also an increase of classification accuracy from Block 0 (no predictors) to the CRI model with the overall correct classification increasing from 84.3% to 88. 9 %. The CRI model accounted for between 2 9.6 % and 51.0 % of the variability in risk status. This model had a negative predictive value of 9 1 2 % and a positive predictive value of 70.8 %. An assessment of the ARI model also revealed it was significantly different from both the original model with no predictors ( 2 (8, N = 217) = 64.72 p < .001) and only using the controlled variables as predictors ( 2 (1, N = 217) = 37.28 p < .001). The final ARI model accounted f or between 2 5 8 % and 44.4 % of the variability in risk status. The correct classification rate rose slightly from 8 4 .3% to 8 8.0 %, and the ARI model had a negative predictive value of 89. 8 % and a positive predictive value of 70.0 %. Both models are summariz ed in Table 18 (Appendix A). Question 5 Differentiation b etween A t risk and non A t risk S tudents What is the optimum number of risk factors for distinguishing between students who are at risk and not at risk? A receiver operating characteristic (ROC) curve was used to determine the optimum number of risk factors which can be used to classify at risk students versus

PAGE 95

85 those students who are not at risk. T his type of analysis calculates the sensitivity and specifi city of each risk factor combination as well as the chances of correctly identifying the risk and non risk groups. An area under the curve of 0.80 or above indicates that the test has good accuracy levels and should also have a s ensitivity above 80 % an d specificity greater than 60 % (Goring, Baldwin, Mariott, Pratt, & Roberts, 2004; Lincoln, Nicholl, Flannaghan, Leona rd, & Van der Gucht, 2003) Table 17 in Appendix A provides a summary of the ROC curve analysis for both the CRI and the ARI. The CRI had an accuracy of 87.7 % correct classification, with a score of 2 or higher resulting in a specificity of 824 and a specificity of 7 43 This means that using 2 as a cut off would correctly identify those students at risk 82.4 % of the time, while incorrectly identifying the non risk students a 25.7 % of the time. Raising the cut off score to 3 would change the sensitivity to 676 and the specificity 913 resulting in 67.6 % of at risk students being captured correctly while incorrectly identifying non risk students 8.7 % of the time. The ROC curve can be seen in Figure 12 (see Appendix A). Looking at the ARI pro duce d results similar to the CRI. Overall, the ARI accurately classified 84.6 % of the students in risk and non risk groups. The cut off score of 2 generated a sensitivity of .7 35 and a specificity of 842 meaning that 7 3 5 % of at risk cases would be id entified correctly and 15.8 % of non risk cases would be identified incorrectly. Raising this score to 1 would increase the sensitivity t o .794 while decreasing the specificity to 672 Figure 13 in Appendix A shows the ROC graph of the ARI.

PAGE 96

86 C hapter Five Discussion The central goal of this study was to develop a more comprehensive school social work model specifically looking at the school domain Providing services with in a host educational setting demands social wo rk services to be focused on areas that impact academic achievement (Lucio, 2008) Using the ecological perspective as a guide, 24 unique school factors were identified which were shown in previous research to impact academic achievement. Fifteen of these factors were analyzed together to get a better understanding of the school domain as it relates to achievement. Specifically a cumulative risk and additive risk approach were used to determine the relationshi p between risk and promotive factors with cumulative grade point average. Findings Initially, fifteen factors had previously been shown to impact achievement related outcomes. F actors were examined individually to determine if they met the criterion for consideration as a risk or promotive factor. Of the original fifteen factors, only five factors met all three criteria to be included i n the CRI and six factors were included in the ARI. Th e se factors were academic expectations (risk and promotive compo nents), academic self efficacy (risk and promotive components) attendance (risk component only), grades repeated (risk only component), music playing (promotive component only), and school behaviors (risk component only)

PAGE 97

87 F our factors met only the first and second criteria as used by others and were eliminated. These included educational support (risk and promotive components), proportion of homework completed (risk only), school belonging (promotive only), and school mobility (risk only). An additiona l analysis was run including these four factors in the CRI and ARI models to determine if there was an improvement in the predictive ability It was found that adding these four factors made the model fit worse for both the CRI and ARI. In fact, t he vari ability in cumulative GPA decreased by 2.9% for the ARI and 3.7% for the CRI with the additional factors. This indicates using only factors that have a unique contribution to the variability of cumulative GPA creates a stronger index than using the previous methods of including factors when they meet the first and second criteria only (Sameroff, 1985) One of the most surprising results was that once the A RI and CRI were added to the models, the effect of SES was no longer significant. This was true for both the standard linear and logistical regression analysis. This suggests the indices accounted for the variability of cumulative GPA that was associated with SES. Model Comparison After determining which factors would be included in each index, the CRI and ARI were compared to determine which model was a better predictor of cumulative GPA The CRI and ARI outcomes were compared using both standard lin ear and logistical regression. Results indicate there is little difference between the two models when using these analyses as a guide. The CRI accounted for less than 1% more variability in GPA when com pared to the ARI. When looking at the logistical r egression outcomes, the CRI accounted for between 4% to 6% more variability in the outcome of cumulative GPA.

PAGE 98

88 While the CRI appears to account for a slightly higher percentage of variability within the current set of predictors, using the ARI is more in line with current standards and views of social work practice. The use of a strengths based perspective is a crucial element of modern social work practice (Breton, 2004; Derezotes, 2000; Miley, O'Melia, & DuBois, 2006; Saleebey, 2005) A strengths or assets based persp ective allows social workers and students to work together to determine an outcome that draws on the (Oko, 2006, p. 602) This approach is built on the notion that in order for any meaningful change to occur, risks as well as promotive factors mus t be included Fraser and Galinsky (2004) stated that a strengths or assets assessed and that assets that may exist in the environment are activated in ways that 394). School social workers must be aware of the challenges that face students as well as the ways protective factors interact to impact student achievement. The National Association of Social Workers (NASW) has set standards for using the strengths ba sed approach in school social work (2002) working with adolescents (2003) and even cultural competence (2001) NASW goes on to say that school social workers s trengths and protective factors can enhance ( 2002, p. 18). Given the slight differences in outcomes between the CRI and the ARI and the alignment of the ARI with the strengths based perspective, the ARI provides more options in reducing risk and enhancing promotive factors. Using this approach is consistent with the work of Dulmus and Rapp Paglicci (2004)

PAGE 99

89 reduced or in some way altered and/or if protective f actors can be enhanced, the likelihood of at risk individuals developing a specific disorder or problem would dec which area the student has a risk component or lacks a pro motive factor. For instance, if a pregnant or parenting student is absent from school 50% of the time, increasing attendance over 80% of the time would increase the likelihood of graduating. Using a program which has been shown to increase attendance rat es of teen mothers would be an appropriate intervention (Harris & Franklin, 2008) By knowing where students have risk or promotive factors in place and where they are lacking will help school social workers intervene in ways that can help students the most. Final Model The results of this study suggest refining the o verall model that was presented in chapter 2 The new model can be seen in F igure 14 in Appendix A and includes only school related factors that have a di rect i mpact on cumulative GPA. The three factors of attendance, grade retention, and school behaviors present as risk only factors. Two factors, academic self efficacy and acad emic expectations, had both risk and protective components, and music playing was found to be promotive only. Putting this domain into context, Figure 15 illustrates the school domain within the framework of the overall school social work model. This mod el presents a guide for social workers to help in determining where interventions would be most effective. In areas that are risk only, moving the student from the risk range would theoretically reduce the potential for failure.

PAGE 100

90 Indices versus D irect F actors The use of an index approach provides greater flexibility to social workers than just examining the direct factors. The results of this dissertation confirm the research of others who have suggested that it is not any specific risk or promotive fa ctor that is as important as the total number of factors (Masten & Powell, 2003; Rutter, 1987; Sameroff & Fiese, 2000; Wachs, 2000) U sing th is approach, if specific r isk factors can be reduced or the prom otive factors can be enhanced there becomes a greater chance of having a higher cumulative GPA. Figure 7 in Appendix A indicates that as the number of risk factors increases, GPA decreases. The same is shown for ARI; as scores increase, GPA also goes up. Factors and Odds of Passing When examining the logistical regression outputs it shows an odds ratio of 2.25, which means that for each addition al point on the additive risk index, by eithe r reducing the risk component or adding to the promotive factors, children have an 125% in creased likelihood of passing When examining the CRI, the odds of passing drop with additional risk. For each risk factor that is added, there is a 72% lower likel ihood of passing. These results show the importance of reducing risk factors and increasing promotive factors. For each risk factor that is removed and each promotive factor that is enhanced, the chances of student success increase tremendously. Cut off P oints The ability of school social workers to be able to determine the optimal point at which students become at most risk of failure is a key component of using an index. If stude nts can be identified at the tipping point of optimal risk, interventions can be

PAGE 101

91 directed at the most needy students. Using either the CRI or ARI would allow school social workers to recognize students at risk and creates opportunities for school social w orkers to intervene appropriately. Looking at the ROC analysis, the optimal cut off point for the CRI is 2 points. Using 2 as a guide, would correctly classify students at risk of failing 82% of the time, while misidentify students as at risk when they a re not 25% of the time. This cutoff point provides the best balance between specificity and sensitivity. Taking this approach with the ARI reveals that with school related factors, using 1 as a cut off would correctly identify students at risk of failure 79% of the time and misclassify those non risk students 32% of the time. C hanging the cutoff to 2 would decrease the miscl assification to 16% of the time, but reduce s correct classification to 74%. While either of these cutoff points would be adequate, a 2 cutoff seems to provide the best balance of correct and incorrect classification. This gives school social workers a solid criterion for being able to identify students at higher risk of failure, and shows where students need to be in order to be mo st successful. Strengths and Limitations This design and implementation of this study attempted to address some of the previous limitations that have been reporte d in other studies. Primarily, the use of school records for grades help ed to strengthen the results that were found. Previous studies have used self report grades which have been shown to be less accurate than school records (Stone & May, 2002) and the use of school records for cumulative GPA certainly makes a stronger case. The second important strength was the number of factors included in the study. While other studies have limited the number of scho ol related factors from one to seven, a total of fifteen school related factors were examined throughout the study.

PAGE 102

92 Two other strengths that are worth mentioning are the sample size and the analysis tools. The sample size of 217 students is well above wh at is suggested as adequate for a regression analysis (Green, 1991) In addition, both linear and logistical regressions were run to compare the results using several techniques The models were also cross validated by splitting the sample and comparing predicted with actual scores. While this study did have numerous strengths, there are several limitations that should be mentioned. Initially it is worth noting the percentage of students in the study with a GPA lower than 2.0 was 15.7%, which is below the 25.1% in the general population. This was th e case even though two drop out prevention schools were selected in order to oversample lower GPA students. It is probable that many students who have lower GPAs may have been missing more school and not received study participation materials during the t ime period they were delivered, already dropped out, or not taken the information home to their parents. Any of these alone would reduce the number of respondents with lower GPAs and taken together could account for the gap between the sample GPA and the overall population GPA. Since only those students with active parental consent and student assent were surveyed, there is no way of knowing which students were not in school or dropped out. In fact, 10.76% of the students were not present during the two week period that the letters and consent forms were sent home Since letters and consent forms were sent home with students at each school, the process relied on students to deliver the items to their parents and then being returned by the students. Whil e there is no way of knowing for sure, it seems likely a high proportion were never even received by parents. In the future, a direct mailing of

PAGE 103

93 information to parents and guardians regarding the study would undoubtedly increase participation rates. A sec ond limitation worth mentioning is that the sample came from a single school district and three schools. While oversampling of lower GPA students was attempted by selecting drop out prevention schools, the sample may not be reflective of other schools or districts around the country Future studies should include a broader selection of schools and students if possible. This would enhance the generalizability of the findings by allowing for the examination of all the school related factors. Future Resea rch Directions While this study has answered many questions regarding school factors and academic achievement, it has also brought to light several areas for future research. The development of the school component is one small aspect of an overall model of school social work. One of the primary next steps will be to begin th e same in depth examination of the other domains of student characteristics, community/neighborhood, family and peers A truly complete model will not be constructed until each of t he five domains is looked at with the same thoroughness as was done with the school factors domain in this dissertation. As each domain is examined it can then be added to an over arching ecological model that can effectively serve school social workers. A second and important area to expand on the results found in this dissertation will be to look at different outcomes in addition to cumulative GPA. Within this dissertation achievement was defined as cumulative GPA, but a chievement can also be viewed as the recent six weeks GPA, math or reading achievement, percentage of credits earned, and even graduation. Each of these outcomes, while within academic

PAGE 104

94 achievement, could be impacted by different factors. It is certainly worth exploring to see whether the factors found here hold up with different outcome measures or if other factors contribute as well. Finally, broadening the s chool factors domain is a logical next step to build on the current work of this dissertation. The two ways to accompli sh this would be to add broader school related factors and to start building the model a step further out. A multi level modeling approach would be useful in identifying the impact of school size, school SES rates, the discipline climate of the school, an d overall school achievement to see what impact these school level factors have on achievement. While taking the model a step further out involves looking at which factors may not directly impact achievement, but whose influence is mediated through anothe r factor. As the model starts to grow further out, the impact of school social workers will also broaden. When looking at attendance, it would be important to know which factors influence attendance so that school social w o rkers can intervene in order t o impact academic achievement. Implications for Social Work Practice The implications of this research carry across all the three domains of ecological systems theory; m acro system, mesosystem, and microsystem. Each of these areas will be discussed in rela tionship to the impact of this dissertation. Macrosystem comprised of values, customs, and laws (Berk, 2006) This dissertation heeds the call of social work for more scientific and evidence based research in social work and social work education (Corcoran, 2007; Shaw, 2003; Zlotnik & Solt, 2006) According to

PAGE 105

95 Pardeck and Yuen (2006) a solid knowledge base is one of the critical components in the social work profession. Fook (2004) goes on t o add that in addition to knowledge development, a research agenda is an additional way that research can contribute in the current climate. In a policy statement regarding school truancy and dropout prevention, NASW (2006) This dissertation enhance d the knowledge base of school social work through the development of an ecological mod el of school social work. The creation of this model was based on previous research and created a more scientific approach to the development o f a school social work model. The introduction of laws that require the use of the Response to Intervention (RtI) model when looking at both general education and special education combines well with the use of a more detailed school social work model. K nowing which factors impact achievement could be strengthened even further if future laws mandated the use of scientifically developed screenings at each tier of intervention Being able to identify students who are at most risk has the potential for inte rv entions to occur in places has the potential improve the chances for academic success. Rather than trying to guess where students need help, a sound schools social work model guides interventions to where they are most needed. In addition to being guided by laws, using this model to guide the delivery of services makes practical sense. The ability to identify students who are at the most risk at each level allows for the most efficient use of time and resources by school social workers. Targeting interv entions to students who need them most enables school social work er s to direct services where they are most needed, optimizing resources. This shift

PAGE 106

96 in thinking continues to build on the path that is being paved with evidence based practice. Mesosystem T he mesosystem is the interaction between two systems, either directly or indirectly. This includes connections between schools and home, and can be cognitive, b ehavior al, or affective (McIntosh, Lyon, Carlson, Everette, & Loera, 2008) Within this context, the support between home and school was measured through home educational support, which was examined directly in the course of this dissertation. In addition, microsystem factors could play a role in the influencing interaction of mesosystem factors (Seginer, 2006) Previous research has shown that parental involvement can be influenced by a number of factors, including school culture (Gardner, Ritb latt, & Beatty, 2000) and student achievement levels (Lewis & Forman, 2002) It is possible that students who have higher absence rates or school behavioral problems could have Further work in this area would be useful in uncoveri ng a complete picture of the relationships that exist between the microsystem structures When looking at the mesosystem, involving families in the education of their children continues to be a vital role for school social workers. Serving as a link be tween home and schools, social workers play a crucial part in this connection. Understanding the factors that help make students successful allows school social workers to partner with families in order to address those areas that present as a risk for ac hievement. Working with families to keep children in school or teaching ways to improve academic self efficacy and expectations strengthens the mesosystem relationship between families

PAGE 107

97 and schools. School social workers work uniquely within the schools a nd with families to generate a partnership that benefits students, schools, and families. It is within this role that understanding a more complete and detailed model of school social work can provide the optimal benefits to impart greater family particip ation. Micro system The most basic level of ecological systems model the microsystem includes relationships and interactions which the student is directly involved Structures include school, family, peers, and neighborhood, and is the closest and most p roximal to the student (Berk, 2006) Within this dissertation six microsystem factors were found to be directly impactful in the school social work model; percent of days present, music playing, academic expectations academic self efficacy, grade retention, and school behaviors. In addition, this research discovered some factors that were previously found to impact academic achievement showed no relationship within the current study. The factors that were not inclu ded in the final model were academic engagement, educational support, percentage of homework completed, school mobility, school safety, school belonging, school relevance, and teacher support. It is certainly plausible that if the outcome of academic achi evement were defined differently, the list of factors showed an impact could change Knowing which factors have a relationship with achievement is important for school social workers. This enables school social workers a place to start in the search for improving academic achievement. Applying this model at a more practical level, i nterventions at the universal level could screen for issues that are school wide This would allow for the implementation of interventions for all students based on an overall need for that particular school Not all

PAGE 108

98 Using this model on a school wide basis would help to identify the unique needs of that specific school. For instance, if a school shows that a majority of the students are at high risk for school behavior problems, school wide interventions could be targeted that that specific risk factor. Moving to a tier 2, or group level, interventions could also be targeted at groups of students who show risk factors, which may not be present in the entire population. Small groups of students could be identified who may have attendance problems, with interventions directed at reducing absences. Finally, for those students where inte rventions were not successful at the first and second tier, could be manage d with individual intercessions. The most intriguing part of applying this model is that the interventions are targeted on areas that have been shown to impact achievement, and cre ate the greatest change of success in the areas of most need. Conclusion have a handle on what is going on in their lives. Bowen and Powers (2005) found that of their students actual experiences matched less than 40% of the time Others have found that school staff had poor knowledge of family functioning (Dwyer, Nicholson, Battistutta, & Oldenburg, 2005) which is supported by Aviles, Anderson, Davila (2006) who found that when students, teachers, and parents were asked to identify the specific reasons for failure, there was little agreement on the causes. All of these findings suggest the need for a quality ecological assessment of school factors as teachers

PAGE 109

99 In order to continue ens uring the social and emotional well being of children it is vital to recognize and identify the multi level risk factors that affect children. Being able to recognize those children who are at the greatest risk due to the presence of mu ltiple risk factor s is an essential task of school social workers. The Council on Social Work (2001) Educational Policy and Accreditation Standards says that content effectively communicate empirically based knowledge supported by the N ational Association of Social Workers Standards for School Social Work Services which notes that school social workers shall be able to select and apply empirically validated or promising preve (p. 20). According to the School Social Work Association of America (Usaj, Shine, & Mandlawitz, 2006) understanding this level of interaction allows school social workers to identify and address systemic barriers to learning. In addition, school social workers can play a crucial role in implementing programs that address and evaluate educational and behavioral concerns. School related factors c an have a large impact on academic achievement over time Konstantantopoulos (2006) found across several studies school factors had a considerable effect on student achievement, consistently accounting for over 50% of the variation in achievement. Th is dissertation developed an index which includes solely school related factors, which enhances t he ability to identify factors which can impede academic achievement This is the first step in being able to intervene and help students succeed. The index developed in this dissertation is a key link between assessing vital factors and being able to select a place to intercede.

PAGE 110

100 As school social workers operate in a system moving toward outcomes and accountability, it is vital to focus attention on how servic es are ultimately impacting the student. The model developed in this dissertation has provided a much more detailed view of the ecological model as it specifically relates to school social work and academic achievement. This will a llow school social work ers to identify pertinent risk and promotive factors that can be addressed in order to reduce the risk of failure by students. The Additive Risk Index and the Cumulative Risk Index provide an opportunity for school social workers to recognize risk factor s that can be addressed through interventions as well as promotive factors that can be expanded or employed. It is the interplay between these factors that creates an overall risk of school failure. It has been shown that the number of risk factors, not any specific risk factor, which is responsible for academic troubles. This give s school social workers the ability to reduce risk and increase protection for struggling students.

PAGE 111

101 References Abu Hilal, M. M. (2000). A structural model of attitudes towards school subjects, academic aspiration and achievement. Educational Psychology, 20 (1), 76 84. Alaimo, K., Olson, C. M., & Frongillo Jr, E. A. (2001). Food insufficiency and American school aged Pediatrics, 108 (1), 44 53. Alliance for Excellent Education. (2003a). FactSheet: The impact of education on crime Washington, DC: Alliance for Excellent Education. Alliance for Excellent Educa tion. (2003b). FactSheet: The impact of education on health & well being Washington, DC: Alliance for Excellent Education. Alliance for Excellent Education. (2003c). FactSheet: The impact of education on personal income & employment Washington, DC: Allia nce for Excellent Education. Alpert, G., & Dunham, R. (1986). Keeping academically marginal youths in school: A prediction model. Youth & Society, 17 (4), 346 361. Alva, S. A., & de Los Reyes, R. (1999). Psychosocial stress, internalized symptoms, and acade mic achievement of Hispanic adolescents. Journal of Adolescent Research, 14 (3), 343 358. Anderman, L. H. (2003). Academic and social perceptions as predictors of change in middle school students' sense of belonging. Journal of Experimental Education, 72 (1) 5 22.

PAGE 112

102 Aviles, A. A., Anderson, T. R., & Davila, E. R. (2006). Child and adolescent social emotional development within the context of school. Child and Adolescent Mental Health, 11 (1), 32 39. Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory Englewood Cliffs, NJ: Prentice Hall. Bartlett, H. (1959). The generic specific concept in social work education and practice In A. E. Kahn (Ed.), Issues in American social work (pp. 159 190). New York: Columbia University Press. Beacon, D. R., & Bean, B. (2006). GPA in research studies: An invaluable but neglected opportunity. Journal of Marketing Education, 28 (1), 35 42. Belsky, J. (1994). The determinants of parenting: A process model. Child Development, 55 83 96. Bergkvist, L. & Rossiter, J. R. (2007). The predictive ability of multiple item versus single item measures of the same constructs. Journal of Marketing Research, 44 175 184. Berk, L. E. (2006). Child development (7th ed.). Boston: Allyn & Bacon. Beverly, S. G. (2001 ). Material hardship in the United States: Evidence from a survey of income and program participation. Social Work Research, 25 (3), 143 151. Bowen, G. L., & Richman, J. M. (2005). The School Success Profile Chapel Hill: Jordan Institute for Families, Scho ol of Social Work, The University of North Carolina at Chapel Hill. Bowen, N. K. (2006). Psychometric properties of the Elementary School Success Profile. Social Work Research, 30 (1), 51 63.

PAGE 113

103 Bowen, N. K., Bowen, G. L., & Ware, W. (2002). Neighborhood socia l disorganization, families, and the educational behavior of adolescents. Journal of Adolescent Research, 17 (5), 468 490. Bowen, N. K., & Powers, J. D. (2005). Knowledge gaps among school staff and the fole of high quality ecological assessments in schools Research on Social Work Practice, 15 (6), 491 500. Breton, M. (2004). An empowerment perspective. In C. D. Garvin, L. M. Gutierrez & M. J. Galinsky (Eds.), Handbook of social work with groups (pp. 58 75). New York: The Guilford Press. Bronfenbrenner, U. (1979). The ecology of human development Cambridge, MA: Harvard University Press. Bronfenbrenner, U. (1989). Ecological systems theory. Annals of Child Development, 6 187 249. Brooks, J. E. (2006). Strengthening resilience in children and youths: Maximizing opportunities through schools. Children & Schools, 28 (2), 69 75. Bruns, E. J., Moore, E., Hoover Stephan, S., Pruitt, D., & Weist, M. D. (2005). The impact of school mental health services on out of school suspension rates. Journal o f Youth and Adolescence, 34 (1), 23 30. Bryant, A. L., Schulenberg, J., Bachman, J. G., O'Malley, P. M., & Johnston, L. D. (2000). Understanding the links among school misbehavior, academic achievement, and cigarette use: A national panel study of adolescen ts. Prevention Science, 1 (2), 71 87.

PAGE 114

104 Burchinal, M., Roberts, J. E., Zeisel, S. A., Hennon, E. A., & Hooper, S. (2006). Social risk and protective child, parenting, and child care factors in early elementary school years. Parenting: Science and Practice, 6 ( 1), 79 113. Bureau of Exceptional Education and Student Services. (2006). The Response to Intervention (RtI) model: Florida Department of Education. Chularut, P., & DeBacker, T. K. (2004). The influence of concept mapping on achievement, self regulation, a nd self efficacy in students of English as a second language. Contemporary Educational Psychology, 29 248 263. Clancy, J. (1995). Ecological social work: The reality and the vision. Social Work in Education, 17 (1), 40 47. Coie, J. D., Miller Johnson, S., & Bagwell, C. (2000). Prevention science. In A. J. Sameroff, M. Lewis & S. Miller (Eds.), Handbook of developmental psychology New York: Plenum Colorado Department of Education. (2005). Fast facts: Evidence Based Practice Retrieved. from http://www.cde.state.co.us/cdesped/download/pdf/ff EvidenceBasedPractice_Intro.pdf Condly, S. J. (2006). Resilience in children: A review of the literature with implications for education. Urban Education, 41 (3), 211 236. Conners, N. A., Bradley, R. H., Mansell, L. W., Liu, J. Y., Roberts, T. J., Burgdorf, K., et al. (2003). Children of mothers with serious substance abuse problems: An accumulation of risks. The American Jour nal of Drug and Alcohol Abuse, 29 (4), 743 758.

PAGE 115

105 Constable, R., & Alvarez, M. (2006). Moving into specialization in school social work: Issues in practice, policy and education. Journal of School Social Work (Special Issue, Summer 2006), 116 131. Constitution of the State of Florida. Article IX. Section 1.7. (2007). Retrieved. from. Cooper, H., Lindsay, J. J., Nye, B., & Greathouse, S. (1998). Relationships among attitudes about homework, amount of homework assigned and completed, and student achievement. Jour nal of Educational Psychology, 90 (1), 70 83. Corcoran, J., Franklin, C., & Bennett, P. (2000). Ecological factors associated with adolescent pregnancy and parenting. Social Work Research, 24 (1), 39 49. Corcoran, K. (2007). From the scientific revolution to evidence based practice: Teaching a short history with a long past. Research on Social Work Practice, 17 (5), 548 552. Corpaci, F. (2008). The role of child temperament on Head Start preschoolers' social compentence in the context of cumulative risk. Journal of Applied Developmental Psychology, 29 (1 16). Costello, E. J., & Agnold, A. (2000). Developmental epidemiology: A framework for developmental psychopathology. In A. J. Sameroff, M. Lewis & S. Miller (Eds.), Handbook of developmental psychopatholog y (pp. 57 73). New York: Plenum. Council on Social Work Education. (2001). Educational policy and accredidation standards : Council on Social Work Education. Crean, H. F., Hightower, A. D., & Allen, M. J. (2001). School based child care for children of teen parents: Evaluation of an urban program designed to keep young others in school. Evaluation and Program Planning, 24 267 275.

PAGE 116

106 Crozier, J. C., & Barth, R. P. (2005). Cognitive and academic functioning in maltreated children. Children & Schools, 27 (4), 197 206. Dandy, J., & Nettelbeck, T. (2002). A cross cultural study of parents' academic standards and educational aspirations for their children. Educational Psychology, 22 (5), 621 627. Davis, L. E., Johnson, S., Miller Cribbs, J., & Saunders, J. (2002). A b rief report: Factors influencing African American youth decisions to stay in school. Journal of Adolescent Research, 17 (3), 223 234. Davis, L. E., Saunders, J., Sharon, J., Miller Cribbs, J., Williams, T., & Wexler, S. (2003). Predicting positive academic intention among African American males and females. Journal of Applied Social Psychology, 33 (11), 2306 2326. Deater Deckard, K. D., K. A., Bates, J. E., & Pettit, G. S. (1998). Multiple risk factors in the development of externalizing behavior problems: Group and individual differences. Development and Psychopathology, 10 (3), 469 493. Derezotes, D. S. (2000). Advanced generalist social work practice Thousa nd Oaks, CA: Sage. DeVellis, R. F. (2003). Scale development: Theory and application (2nd ed. Vol. 26). Thousand Oaks, CA: Sage Publications. Dimmitt, C. (2003). Transforming School Counseling Practice Through Collaboration and the Use of Data: A Study of Academic Failure in High School. Professional School Counseling, 6 (5), 340 349.

PAGE 117

107 Dinkes, R., Cataldi, E. F., & Kena, G. (2006). Indicators of School Crime and Safety: 2006 (NCES 2007 003/NCJ 214262) Washington, DC: U.S. Government Printing Office. Dulmus, C. N., & Rapp Pagglicci, L. A. (2004). Prevention and resilience. In L. A. Rapp Pagglicci, C. N. Dulmus & J. S. Wodarski (Eds.), Handbook of Prevention Interventions for Children and Adolescents Hoboken, NJ: John Wiley & Sons, Inc. Dunn, M. C., Kadane, J. B., & Garrow, J. R. (2003). Comparing harm done by mobility and class absence: Missing students and missing data. Journal of Educational Behavior and Statistics, 28 (3), 269 288. Dupper, D. R. (2003). School social work: Skills and interventions for effect ive practice Hoboken, NJ: John Wiley & Sons, Inc. Dwyer, S. B., Nicholson, J. M., Battistutta, D., & Oldenburg, B. (2005). Teachers' knowledge of children's exposure to family risk factors: Accuracy and usefulness. Journal of School Psychology, 43 23 38. Eamon, M. K. (2002). Effects of poverty on mathematics and reading achievement of young adolescents. Journal of Early Adolescence, 22 (1), 49 74. Eckert, T. L., Dunn, E. K., Codding, R. S., Begeny, J. C., & Kleinmann, A. E. (2006). Assessment of mathematic s and reading performance: An examination of the correspondance between direct assessment of student performance and teacher report. Psychology in the Schools, 43 (3), 247 265. Evans, S. W. (1999). Mental health services in schools: Utilization, effectivene ss, and consent. Clinical Psychology Review, 19 (2), 165 178.

PAGE 118

108 Ezpelata, L., Granero, R., de la Osa, N., & Domenech, J. M. (2008). Risk factor clustering for psychopathology in socially at risk Spanish children. Social Psychiatry and Psychiatric Epedimiology 43 559 568. Finn, J. D. (2006). The adult lives of at risk students: The roles of attainment and engagement in high school (No. NCES 2006 328). U.S. Department of Education, Washington, D.C.: National Center for Educational Statistics. Fitzpatrick, K. R (2006). The effect of instrumental music participation and socioeconomic status on Ohio fourth sixth and ninth grade proficiency test performance. JRME, 54 (1), 73 84. Florida Senate. (2007). The 2007 Florida Statutes, K 12 Education Code:1003.43. http://www.flsenate.gov/ Fook, J. (2004). What professionals need from research: Beyond evidence based practice. In D. Smith (Ed.), Research highlights in social work: Social work and evidence based practice (pp. 29 46) Philadelphia: Jessica Kingsley Publishers. Fraser, M. W. (2004). The ecology of childhood: A multisystems perspective. In M. W. Fraser (Ed.), Risk and resilience in childhood (2nd ed., pp. 1 12). Washington, D.C.: NASW Press. Fraser, M. W., & Galinsky, M J. (2004). Risk and resilience in childhood: Toward an evidence based model of practice. In M. W. Fraser (Ed.), Risk and Resilience in Childhood (2nd ed., pp. 385 402). Washington, DC: NASW Press. Fraser, M. W., Kirby, L. D., & Smokowski, P. R. (2004). R isk and resilience in childhood. In M. W. Fraser (Ed.), Risk and resilience in childhood: An ecological perspective (2nd ed., pp. 13 66). Washington, DC: NASW Press.

PAGE 119

109 Fraser, M. W., Richman, J. M., & Galinsky, M. J. (1999). Risk, protection, and resilience: Toward a conceptual framework for social work practice. Social Work Research, 23 (3), 131 143. Furstenberg, F. E., Cook, T., Eccles, J. S., Elder, G. H., & Sameroff, A. J. (1999). Urban families and adolescent success Chicago: University of Chicago Press. Gambone, M. A., Klem, A. M., & Connell, J. P. (2002). Finding out what matters for youth: Testing key links in a community action framework for youth development Philadelphia, P.A.: Youth Development Strategies, Inc. and Institute for Research and Reform in Education. Gardner, P. W., Ritblatt, S. N., & Beatty, J. R. (2000). Academic achievement and parental involvement as a function of high school size. High School Journal, 83 21 27. Garmezy, N., & Masten, A. S. (1994). Chonic adversities. In M. Rutter, L. Herzov & E. Taylor (Eds.), Child and adolescent psychiatry (pp. 191 208). Oxford: Blackwell Scientific. Garrett, K. J. (2007a). Ecological persepctive for school social work practice. In L. Bye & M. Alvarez (Eds.), School Social Work: Theory to Practice Belmont, CA: Thompson Brooks/Cole. Garrett, K. J. (2007b). Ecological perspective for school social work practice. In L. Bye & M. Alvarez (Eds.), School Social Work: Theory to Practice Belmont, CA: Thompson Brooks/Cole.

PAGE 120

110 Gassman Pines, A., & Yoshikawa, H. (2006). The effects of antipoverty programs on related risk. Developmental Psychology 42 (6), 981 999. Germain, C. B. (2006). An ecological perspective on social work in the schools. In R. Constable, C. R. Massatt, S. McDonald & J. P. Flynn (Eds.), School social work: Practice, policy, and research (6th ed., pp. 28 39). Chicago, IL: Lyceum Books, Inc. Glover, T. A., & Albers, C. A. (2007). Considerations for evaluating universal screening assessments. Jou rnal of School Psychology, 45 117 135. Gonzalez Pienda, J. A., Nunez, C. J., Gonzalez Pumariega, S., Alvarez, L., Roces, C., & Garcia, M. (2002). A structural equation model of parental involvement, motivational and aptitudinal characteristics, and academ ic achievement. Journal of Experimental Education, 70 (3), 257 287. Goodenow, C. (1993). Classroom belonging among early adolescent students: Relationships to motivation and achievement. Journal of Early Adolescence, 19 (3), 21 43. Gorard, S., Rees, G., & Sa lisbury, J. (2001). Investigating the patterns of differential attainment of boys and girls at school. British Educational Research Journal, 27 (2), 125 139. Gordon, R. (1987). An operational classification of disease and prevention. In J. A. Steinberg & M. M. Silverman (Eds.), Preventing mental disorders: A research perspective (pp. 20 26). Rockville, MD: National Institute of Mental Health.

PAGE 121

111 Goring, H., Baldwin, R., Mariott, A., Pratt, H., & Roberts, C. (2004). Validation of short screening tests for depres sion and cognitive impairment in older medically ill inpatients. International Journal of Geriatric Psychiatry, 19 (465 471). Green, S. B. (1991). How many subjects does it take to do a regression analysis? Multivariate Behavioral Research, 26 (3), 499 510. Grinnell, R. M., Unrau, Y. A., & Williams, M. (2005). Group level Designs. In R. M. Grinnell & Y. A. Unrau (Eds.), Social Work Research and Evaluation: Quantitative and Qualitative Approaches (7th ed.). New York: Oxford University Press. Gutman, L. M., & M cLoyd, V. C. (2000). Parents' management of their children's education within the home, at school, and in the community: An examination of African American families living in poverty. Urban Review, 32 (1), 1 24. Gutman, L. M., & Midgley, C. (2000). The role of protective factors in supporting the academic achievement of poor African American students during the middle school transition. Journal of Youth and Adolescence, 29 (2), 223 248. Gutman, L. M., Sameroff, A. J., & Cole, R. (2003). Academic growth curve trajectoiries from 1st grade to 12th grade: Effects of multiple social risk factors and preschool child factors. Developmental Psychology, 39 (4), 777 790. Gutman, L. M., Sameroff, A. J., & Eccles, J. S. (2002). The academic achievement of African American students during early adolescence: An examination of multiple risk, promotive, and protective factors. American Journal of Community Psychology, 30 (3), 367 400.

PAGE 122

112 Hamre, B. K., & Pianta, R. C. (2001). Early teacher child relationships and the trajectory of c hildren's school outcomes through eighth grade. Child Development, 72 (2), 625 638. Harlow, C. W. (2003). Education and correctional populations Washington, DC: U.S. Department of Justice. Harris, M. B., & Franklin, C. (2008). Taking charge: A school based life skills program for adolescent mothers New York: Oxford University Press. Henderson, A. T., & Mapp, K. L. (2002). A New wave of evidence: The impact of school, family, and community connections on student achievement Austin, TX: National Center for Family & Community Connections with Schools. Jenson, J. M. (2004). Risk and protective factors for alcohol and other durg use in chilhood and adolescence. In M. W. Fraser (Ed.), Risk and resilience in childhood: An ecological perspective (2nd ed., pp. 183 208). Washington, DC: NASW Press. Jimerson, S. (1999). On the failure of failure: Examining the association between early grade retention and education and employment outcomes during late adolescence. Journal of School Psychology, 37 (243 272). Jimerson, S. Anderson, G. E., & Whipple, A. D. (2002). Winning the battle and losing the war: Examining the relation between grade retention and dropping out of school. Psychology in the Schools, 39 (4). Jimerson, S., Carlson, E., Rotert, M., Egeland, B., & Sroufe, L. A. (1997). A prospective, longitudinal study of the correlates and consequences of early grade retention. Journal of School Psychology, 35 (1), 3 25.

PAGE 123

113 Jimerson, S. R., Egeland, B., Sroufe, L. A., & Carlson, B. (2000). A prospective longitudinal study of hig h school dropouts: Examining multiple predictors across development. Journal of School Psychology, 38 (6), 525 549. Johnson, J. A., Biegel, D. E., & Shafran, R. (2000). Concept mapping in mental health: Uses and adaptations. Evaluation and Program Planning, 23 67 75. Johnsson, E., & Svensson, K. (2005). Theory in social work: Some reflections on understanding and explaining interventions. European Journal of Social Work, 8 (4), 419 433. Jones, K. (2004). Assessing psychological separation and academic perfor mance in nonresident father and resident father adolescent boys. Child and Adolescent Social Work Journal, 21 (4), 333 354. Keith, D. (1989). Refining concept maps: Methodologicial issues and an example. Evaluation and Program Planning, 12 (1), 75 80. Kellow, J. T., & Jones, B. D. (2008). The effects of stereotypes on the achievement gap: Reexamining the academic performance of African American high school students. Journal of Black Psychology, 34 (1), 94 120. Kennedy, A. C., & Bennett, L. (2006). Urban adolescent mothers exposed to community, family, and partner violence. Journal of Interpersonal Violence, 21 (6), 750 773. Kerr, M. A., Black, M. M., & Krishnakumar, A. (2000). Failure to thrive, maltreatment and the behavior and development of 6 year old c hildren from low income, urban families: A cumulative risk model. Child Abuse & Neglect, 24 (5), 587 598. Kinard, E. M. (2001). Perceived and actual academic competence in maltreated children. Child Abuse & Neglect, 25 33 45.

PAGE 124

114 Klem, A. M., & Connell, J. P. (2004). Relationships matter: Linking teacher support to student engagement and achievement. Journal of School Health, 74 (7), 262 273. Konstantopoulos, S. K. (2006). Trends of School Effects on Student Achievement: Evidence from NLS:72, HSB:82, and NELS:92 Teachers College Record, 108 (12), 2550 2581. Kuncel, N. R., Crede, M., & Thomas, L. L. (2005). The Validity of self reported grade point averages, class ranks, and test scores: A meta analysis and review of the literature. Review of Educational Research, 75 (1), 63 82. Laird, J., DeBell, M., & Chapman, C. (2006). Dropout rates in the United States: 2004, NCES 2007 024 Washington, DC: National Center for Educational Statistics. Lapan, R. T., Gysbers, N. C., & Petroski, G. F. (2001). Helping seventh graders be safe and successful: A statewide study of the impact of comprehensive guidance and counseling programs. Journal of Counseling & Development, 79 (3), 320 330. Lee, W. C. (2007). The many facets in the roleof a school social worker. In L. Bye & M. Alvarez (Eds.), School social work: Theory to practice (pp. 51 65). Belmont, CA: Thompson Brooks/Cole. Lewis, A. E., & Forman, T. A. (2002). Contestation or collaboration? A comparative study of home school relations. Anthropology & Education, 33 60 89. Lincoln, N. B., Nicholl, C. R., Flannaghan, T., Leonard, M., & Van der Gucht, E. (2003). The validity of questionnaire measures for assessing depression after stroke. Clinical Rehabilitation, 17 840 846. Loo, R. (2002). A caveat on using single item versus multip le item scales. Journal of Managerial Psychology, 17 (1/2), 68 75.

PAGE 125

115 Lucio, R. (2006). A review of the literature on theories on adolescent fatherhood. Journal of School Social Work, 13 2 12. Lucio, R. (2008). Bridging Theory, Research and Roles in School So cial Work: A New Paradigm. Unpublished manuscript sumbitted for publication. Luthar, S. S. (1991). Vulnerability and resilience: A study of high risk adolescents. Child Development, 62 600 616. Luthar, S. S., Cicchetti, D., & Becker, B. (2000). Research on resilience: A critical evaluatoin and guidelines for future work. Child Development, 71 573 575. Luthar, S. S., & Zelazo, B. L. (2003). Research on resilience: An integrative review. In S. S. Luthar (Ed.), Resilience and vulnurability: Adaptation in th e context of childhood adversitites (pp. 511 549). New York: Cambridge University Press. Lynn, C. J., McKay, M. M., & Atkins, M. S. (2003). School social work: Meeting the metnal health needs of students through collaboration with teachers. Children & Scho ols, 25 (4), 197 209. Ma, X., & Klinger, D. A. (2000). Hierarchical linear modelling of student and school effects on academic achievement. Canadian Journal of Education, 25 (1), 41 55. Magdol, L. (1994). Risk factors for adolescent academic achievement Mad ison, WI: University of Wisconsin Madison Cooperative Extension. Marchant, G. J., Paulson, S. E., & Rothlisberg, B. A. (2001). Relations of middle school students' perceptions of family and school contexts with academic achievement. Psychology in the Schoo ls, 38 (6), 505 519.

PAGE 126

116 Marks, H. (2000). Student engagement in instructional activity: Patterns in elementary, middle, and high school years. American Educational Research Journal, 37 (1), 153 184. Marsh, H. W., Martin, A. J., & Cheng, J. H. S. (2008). A multi level perspective on gender in classroom motivation and climate: Potential benefits of male teachers for boys? Journal of Educational Psychology, 100 (1), 78 95. Masten, A. S., & Coatsworth, J. D. (1998). The development of competence in favorable and unfa vorable environments: Lessons from research on successful children. American Psychologist, 53 205 220. Masten, A. S., & Obradovic, J. (2006). Competence and resilience in development. In B. M. Lester, A. S. Masten & B. McEwen (Eds.), Resilience in childre n (Vol. 1094, pp. 13 27). Boston: Blackwell Publishing. Masten, A. S., & Powell, J. L. (2003). A resilience framework for research, policy, and practice. In S. S. Luthar (Ed.), Resilience and vulnerability: Adaptation in the context of childhood adversitie s (pp. 1 25). New York: Cambridge University Press. McIntosh, J. M., Lyon, A. R., Carlson, G. A., Everette, C. D. B., & Loera, S. (2008). Measuring the mesosystem: A survey and critique of approaches to cross setting measurement for ecological research and models of collaborative care. Families, Systems, & Health, 26 (1), 86 104. McMurty, S. L. (2005). Surveys. In R. M. Gri nnell & Y. A. Unrau (Eds.), Social Work Research and Evaluation: Quantitative and Qualitative Approaches (7th ed.). New York: Oxford University Press.

PAGE 127

117 Mehana, M., & Reynolds, A. J. (2004). School mobility and achievement: A meta analysis. Children and Yout h Services Review, 26 93 119. Menning, C. L. (2006). Non resident fathering and school failure. Journal of Family Issues, 27 (10), 1356 1382. Miley, K. K., O'Melia, M. W., & DuBois, B. L. (2006). Generalist social work practice (5th ed.). Boston: Allyn & Bacon. Miller, D. B., & MacIntosh, R. (1999). Promoting resilience in urban African American adolescents: Racial socialization and identity as protective factors. Social Work Research, 23 (3), 159 170. Mitchell, S. K., Bee, H. L., Hammond, M. A., & Barna rd, K. E. (1985). Predictions of school behavior and problems in chidlren frollowed from birth to age eight. In W. K. Frankeburg, R. N. Emde & J. W. Sullivan (Eds.), Early identification of children at risk: An international perspective (pp. 117 132). New York: Plenum Press. Moller, S., Stearns, E., Blau, J. R., & Land, K. C. (2006). Smooth and rough roads to academic achievement: Retention and race/class disparities in high school. Social Science Research, 35 157 180. Muller, C. (2001). The role of caring in the teacher student relationship for at risk students. Sociological Inquiry, 71 (2), 241 255. Mullis, R. L., Rathge, R., & Mullis, A. K. (2003). Predictors of academic performance during early adolescence: A contextual view. International Journal of Beh avioral Development 27 (6), 54 548.

PAGE 128

118 Murray, C., & Malmgren, K. (2005). Implementing a teacher student relationship program in a high poverty urban school: Effects on social, emotional, and academic adjustment and lessons learned. Journal of School Psycholog y, 43 137 152. National Association of Social Workers. (2001). NASW standards for cultural competence in social work practice Washington, DC: NASW. National Association of Social Workers. (2002). NASW standards for school social work services Washington DC: NASW. National Association of Social Workers. (2003). NASW standards for the practice of social work with adolescents Washington, DC: NASW. National Association of Social Workers. (2006). Social work speaks: National Association of Social Workers po licy statements 2006 2009 Washington, DC: NASW Press. National Research Council. (2004). Engaging schools: Fostering high school students' motivation to learn Washington, D.C.: National Academy Press. No Child Left Behind Act of 2001. Pub. L. No. 107 110. Nunnally, J. C. (1978). Psychometric theory (2nd ed.). New York: McGraw Hill. Obach, M. S. (2003). A longitudinal sequential study of perceived academic competence and motivational beliefs for learning among children in middle school. Educational Psycholo gy, 23 (3), 323 338. Oko, J. (2006). Evaluating alternative approaches to social work: A critical view of the strengths perspective. Families in Society, 87 (4), 601 611.

PAGE 129

119 Ong, A. D., Phinney, J. S., & Dennis, J. (2006). Competence under challenge: Exploring the protective influence of parental support and ethnic identity in Latino college students. Journal of Adolescence, 29 961 979. Pardeck, J. T., & Yuen, F. K. O. (2006). Social work for the twenty first century: Challenges and opportunities Westport, CT: Praiger. Perie, M., Grigg, W., & Dion, G. (2005). (NCES 2006 453). U.S. Department of Education, National Center for Education Statistics. Washington, D.C.: U.S. Government Printing Office. Powell, C. L., & Arrio la, K. R. J. (2003). Relationship between psychosocial factors and academic achievement among African American students. Journal of Educational Research, 96 (3), 175 181. Powers, J. D., Bowen, G. L., & Rose, R. A. (2005). Using social environment assets to identify intervention strategies for promoting school success. Children & Schools, 27 (3), 177 187. Prelow, H. M., & Loukas, A. (2003). The role of resource, protective, and risk factors on academic achievement related outcomes of economically disadvantaged Latino youth. Journal of Community Psychology, 31 (5), 513 529. Richman, J. M., Bowen, G. L., & Woolley, M. E. (2004). School failure: An eco interactional developmental perspective. In M. W. Fraser (Ed.), Risk and resilience in childhood: An ecological pe rspective (2nd ed., pp. 133 160). Washington, DC: NASW Press. Roscigno, V. J. (2000). Family/school inequality and African American/Hispanic achievement. Social Problems, 47 (2), 266 290.

PAGE 130

120 Rubin, A., & Babbie, E. (2004). Research Methods for Social Work (5th ed.). New York: Wadsworth. Rumberger, R. W., & Palardy, G. J. (2005). Does segregation still matter? The impact of student composition on academic achievement in high school. Teachers College Record, 107 (9), 1999 2045. Rutter, M. (1979). Protective f actors in children's response to stres and disadvantage. In M. W. Kent & J. E. Rolf (Eds.), Primary prevention in psychopathaology: Social competence in children (Vol. 3). Hanover, NH: University Press of New England. Rutter, M. (1987). Psychosocial resili ence and protective mechanism. American Journal of Orthopsychiatry, 57 316 331. Rutter, M. (2000). Resilience reconsidered: Conceptual considerations, empirical findings, and policy implications. In J. P. Shonkoff & S. J. Meisels (Eds.), Handbook of early childhood intervention (2 ed., pp. 651 682). New York: Cambridge University Press. Rutter, M. (2006). Implications of resilience concepts for scientific understanding. In B. M. Lester, A. S. Masten & B. McEwen (Eds.), Resilience in children (Vol. 1094, pp 1 12). Boston: Blackwell Publishing. Saleebey, D. (2005). The strengths perspective in social work practice (4th ed.). Boston: Allyn & Bacon. Sameroff, A. J. (1985). Environmental factors in the early screening of children at risk. In W. K. Frankeburg, R N. Emde & J. W. Sullivan (Eds.), Early identification of children at risk: An international perspective (pp. 21 45). New York: Plenum Press.

PAGE 131

121 Sameroff, A. J. (2003). Ecological perspectives on development risk. In J. D. Osofsky & H. E. Fitzgerald (Eds.), WAIMH handbook of infant mental health (Vol. 4, pp. 1 33). New York: John Wiley & Sons, Inc. Sameroff, A. J. (2006). Identifying risk and protective factors for healthy child development. In A. Clarke Stewart & J. F. Dunn (Eds.), Families count: Effects on child and adolescent development (pp. 53 78). New York: Cambridge University Press. Sameroff, A. J., & Fiese, B. H. (2000). Transactional regulation: The developmental ecology of early intervention. In J. P. Shonkoff & S. J. Meisels (Eds.), Handbook of ea rly childhood intervention (2nd ed., pp. 135 159). New York: Cambridge University Press. Sameroff, A. J., Gutman, M. L., & Peck, S. C. (2003). Adaptations among youth facing multiple risks: Prospective research findings. In S. S. Luthar (Ed.), Resilience a nd vulnerability: Adaptation in the context of childhood adverstiy (pp. 364 391). New York: Cambridge University Press. Sameroff, A. J., Seifer, R., Zax, M., & Barocas, R. (1987). Early indicators of developmental risk: The Rochester Longitudinal Study. Sc hizophrenia Bulletin, 13 383 393. Sanchez, B., Colon, Y., & Patricia, E. (2005). The role of sense of school belonging and gender in the academic adjustment of Latino adolescents. Journal of Youth and Adolescence, 34 (6), 619 628. Sanders, C. E., Field, T. M., & Diego, M. A. (2001). Adolescents' academic expectations and achievement. Adolescence, 36 (144), 795 802.

PAGE 132

122 Saunders, J., Davis, L., Williams, T., & Williams, J. H. (2004). Gender differences in self perceptions and academic outcomes: A study of African American high school students. Journal of Youth and Adolescence, 33 (1), 81 90. Schellenberg, E. G. (2004). Music lessons enhance IQ. Psychological Science, 15 (8), 511 514. Schellenberg, E. G. (2006). Long term positive associations between music lessons a nd IQ. Journal of Educational Psychology, 98 (2), 457 468. perspective. Parenting: Science and Practice, 6 (1), 1 48. Shaw, I. F. (2003). Critical commentary: Cutting edge issues i n social work research. British Journal of Social Work, 33 107 116. Shern, D. L., Trochim, W. M. K., & LaComb, C. A. (1995). The use of concept mapping for assessing fidelity of model transfer: An Example From Psychiatric Rehabilitation. Evaluation and Pr ogram Planning, 18 (2), 143 153. Sink, C. A., & Stroh, H. R. (2003). Raising achievement test scores of early elementary school students through comprehensive school counseling programs. Professional School Counseling, 6 (5), 350 364. Sirin, S. R., & Rogers Sirin, L. (2004). Exploring school engagement of middle class African American adolescents. Youth & Society, 35 (3), 323 340. Stone, C. A., & May, A. L. (2002). The accuracy of academic self evaluations in adolescents with learning disabilities. Journal of Learning Disabilities, 35 (4), 370 383.

PAGE 133

123 Suldo, S. M., & Shaffer, E. J. (2007). Evaluation of the self efficacy questionnaire for children in two samples of American adolescents. Journal of Psychoeducational Assessment, 25 (4), 341 355. Swanon, C. B., & Schneider, B. (1999). Students on the move: Residential and educational mobility in America's schools. Sociology of Education, 72(January) 54 67. Tavani, C. M., & Losh, S. C. (2003). Motivation, self confidence, and expectations as predictors of the academic performances among our high school students. Child Study Journal, 33 (3), 141 151. Teasley, M. L., & Lee, E. (2006). Examining the association between academic achievement and self esteem in African American male youth in a community outreach after school program. School Social Work Journal, 30 (2), 64 81. Thomlison, B. (2004). Child maltreatment: A risk and protective factor perspective. In M. W. Fraser (Ed.), Ri sk and resilience in childhood: An ecological Perspective (2 ed., pp. 89 132). Washington, DC: NASW Press. Thorstensen, B. I. (2004). If you build it, they will come: Investing in public education. Retrieved 10/01/07, from http://abec.unm.edu/resources/gallery/present/invest_in_ed.pdf Trochim, W. M. K. (1989a). An introduction to concept mapping for program evaluation. Evaluation and Program Planning, 12 1 16. Trochim, W. M. K. (1989b). Concept mapping: Soft science or hard art? Evaluation and Program Planning, 12 87 110.

PAGE 134

124 U.S. Department of Edcuation, National Center for Educational Statistics. (2006). The condition of education 2006, NCES 20006 076 Washington, DC: US Governmen t Printing Office. Usaj, K., Shine, J. K., & Mandlawitz, M. (2006). New roles in response to intervention: New roles for school social workers : School Social Work Association of America. Wachs, T. D. (2000). Necessary but not sufficient Washington, DC: Am erican Psychological Association. Wakefield, J. C. (1996a). Does social work need the eco systems perspective (Part 1): Is the perspective clinically useful. Social Service Review, 70 (1), 1 32. Wakefield, J. C. (1996b). Does social work need the eco system s perspective (Part 2): Does the perspective save social work from incoherence. Social Service Review, 70 (2), 183 213. Werner, E. E. (2000). Protective factors and individual resilience. In J. P. Shonkoff & S. J. Meisels (Eds.), Handbook of early childhood edition (2 ed., pp. 115 132). New york: Cambridge University Press. Williams, J. H., Ayers, C. D., Van Dorn, R. A., & Arthur, M. W. (2004). Risk and protective factors in the developmental of delinquency and conduct disorder. In M. W. Fraser (Ed.), Risk a nd resilience in childhood: An ecological perspective (2nd ed., pp. 209 250). Washington, DC: NASW Press. Woods, G. E. (1994). School improvement research series: Reducing the drop out rate. Retrieved 10/01/07, from http://www.nwrel.org/scpd/sirs/9/c017.html Woolley, M. E., & Bowen, G. L. (2007). In the context of risk: Supportive adults and the school engagement of middle school students. Family Relations, 56 92 104.

PAGE 135

125 Woolley, M. E., & Grogan Kaylo r, A. (2006). Protective family factors in the context of neighborhood: Promoting positive school outcomes. Family Relations 55 (1), 93 104. Wyman, P. A. (2003). Emerging perspectives on context specificity of children's adaptation and resilience: Evidence from a decade of research with urban children in adversity. In S. S. Luthar (Ed.), Resilience and vulnerability: Adaptation in the context of childhood adversities (pp. 293 317). New York: Cambridge University Press. Yates, T. M., Egeland, B., & Sroufe, L A. (2003). Rethinking resilience: A developmental process prospective. In S. S. Luthar (Ed.), Resilience and vulnerability: Adaptation in the context of childhood adversities (pp. 243 265). New York: Cambridge University Press. Zand, D. H., & Thomson, N. R. (2005). Academic achievent among African American adolescents: Direct adn indirect effects of demographic, individual and contextual variables. Journal of Black Psychology, 31 (4), 352 368. Zimmerman, B. J., Bandura, A., & Martinez Pons, M. (1992). Self motivation for academic attainment: The role of self efficacy beliefs and personal goal setting. American Educational Research Journal, 29 (3), 663 676. Zlotnik, J. L., & Solt, B. E. (2006). The Institute for the Advancement of Social Work Research: Workin g to increase our practice and policy evidence base. Research on Social Work Practice, 16 (5), 534 540.

PAGE 136

126 Appendices

PAGE 137

127 Appendix A : Tables and Figures Figure 1. Broad Ecological Model

PAGE 138

128 Appendix A (Continued) Table 1. Articles E xamined for L iterature R eview All Years Academic Achievement Educational Outcomes Combined Social Work Abstract Plus 297 30 327 PSYCINFO 39, 395 702 40,103 ERIC 56,386 8,920 65,306 Total 9 6,078 9,652 105,740 2000 2007 Social Work Abstracts Plus 103 12 115 PSYCINFO 9,602 392 9,994 ERIC 8,099 2,511 10,610 Total 17,804 2,915 20,719

PAGE 139

129 Appendix A (Continued) Table 2. Factors R elated to A cademic A chievement 1 Absent Parent Involvement 41 Family Member Dropped Out 81 Peer Academic Performance 2 Academic Engagement 42 Foster Care or Public Care 82 Peer Acceptance 3 Academic Self Efficacy 43 Gay, Lesbian, Bisexual 83 Peer Behaviors 4 Achievement Gap 44 Gender 84 Peer Religion 5 Adoption 45 Grade Retention 85 Peer Support 6 After school programs 46 Growth and Development 86 Personality Type 7 Age of Mother at Birth 47 Home Educational Resou r ces 87 Physical Health Status 8 Aggression 48 Homelessness 88 Pro Social Behaviors 9 Anxiety 49 Homework 89 Relationships and Dating 10 Assets/Income 50 Immigrant proportion 90 Religion Family 11 Attendance 51 IQ 91 Religiosity Religion 12 Attitude towards school 52 Learning Disability 92 Residential Father 13 Basic Needs 53 Life Stressors 93 School Behaviors 14 Birth Order 54 Locus of Control 94 School belonging 15 Birth Season 55 Maltreatment/Abuse 95 School District Size 16 Birth Weight 56 Maternal Depression 96 School Minority Rates 17 Birthday 57 Maternal Employment 97 School Quality 18 Breast Feeding 58 Maternal Health 98 School Relevance 19 Bullying 59 Mental Health Factors 99 School Safety 20 Child Support 60 Mentors 100 School SES 21 Class Size 61 Motivation 101 School Size 22 Counseling 62 Music Instruction 102 School/Residential mobility 23 Crime 63 Neighborhood Characteristics 103 Self Esteem 24 Cultural Factors 64 Neighborhood Deterioration 104 Self Regulated Learning 25 Daycare and preschool 65 Neighborhood Quality 105 Sleep Time 26 Delinquency 66 Neighborhood resources 106 Social Competence 27 Economic Status or Poverty 67 Neighborhood Violence 107 Social Skills 28 Educational Support 68 Neighborhood Behaviors 108 Social Support 29 Emotional/Behav. Disorders 69 Number of Siblings 109 Student Expectations 30 Employment 70 110 Student Smoking 31 English Fluency ESL 71 Parental Expectations 111 Student Substance Use 32 Ethnic Identity 72 Parental Distress 112 Supervision of Homework 33 Extra Curricular Activities 73 Parental Education 113 Supportive Adults 34 Family Cohesion 74 Parental Involvement 114 Supportive School Environ. 35 Family Comp. of School 75 Parental Monitoring 115 Teacher Relationships 36 Family Management 76 Parental School Involvement 116 Teacher Support 37 Family Stressful Events 77 Parental Child Attachment 117 Television and Computer Use 38 Family Structure (1 parent) 78 Perceived safety 118 Urbanicity (Rural/Urban/City) 39 Family Support 79 Parent child conflict 40 Family Trauma 80 Parenting Style

PAGE 140

130 Appendix A (Continued) Table 3. Factors Above 75% Initial Agreement Domain Factor % Domain Factor % F Family Cohesion 100.0 P Peer Behaviors 89.7 N Neighborhood Resources 100.0 S School Quality 89.7 N Crime (Neighborhood) 98.3 S School SES 89.7 F Family Management 98.3 S Teacher Relationships 89.7 N Neigh. Characteristics 98.3 F Child Support 87.9 N Neigh. Deterioration 98.3 C Gay, Lesbian, Bisexual 87.9 N Neighborhood Quality 98.3 C Gender 87.9 F Religion Family 98.3 P Peer Religion 87.9 S School Safety 98.3 P Peer Support 87.9 F Assets/Income 96.6 F Poverty or Economic Status 87.9 S Class Size 96.6 S School Size 87.9 F Family Structure (1 parent) 96.6 F Television /Computer Use 87.9 F Maternal Depression 96.6 F Family Dropped Out 86.2 F Maternal Employment 96.6 F Maltreatment/Abuse 86.2 F Parental Distress 96.6 F Parental Education 86.2 F Parenting Style 96.6 F Economic Status 84.5 F Supervision of Homework 96.6 F Parental Involvement 84.5 S Supportive School Environ 96.6 P Peer Acceptance 84.5 S Teacher Support 96.6 S School District Size 84.5 F Number of Siblings 94.8 C Anxiety 82.8 F Absent Parent Involvement 93.1 F Breast Feeding 82.8 N Neighborhood Violence 93.1 N Immigrant % Community 82.8 F Parental Child Attachment 93.1 S School Relevance 82.8 F Parent Child Conflict 93.1 F Family Support 81.0 C Personality Type 93.1 S Grade Retention 81.0 F Basic Needs 91.4 N Neigh. Youth Behaviors 81.0 F Family Stressful Events 91.4 S School Belonging 79.3 F Maternal Health 91.4 S School Minority Rates 79.3 F Parent Academic Expect. 91.4 S Academic Engagement 77.6 F Parental Monitoring 91.4 C Learning Disability 77.6 F Residential Father 91.4 C Self Esteem 77.6 N Urbanicity 91.4 C Self Regulated Learning 77.6 F Age of Mother at Birth 89.7 C Aggression 75.9 F Family Trauma 89.7 F Ethnic Identity 75.9 C IQ 89.7

PAGE 141

131 Appendix A (Continued) Figure 2 Multi dimensional S caling P lot

PAGE 142

132 Appendix A (Continued) Table 4. Doma i n and I ndividual F actors Child Aggression Gender Pro Social Behaviors Anxiety Growth and Development Self Esteem Attitude Towards School IQ Self Regulated Learning Birth Order Learning Disability Sleep Time Birth Season Locus of Control Social Competence Birth Weight Mental Health Factors Social Skills Birthday Motivation Student Smoking Emotional/Behav. Disorders Personality Type Student Substance Use Gay, Lesbian, Bisexual Physical Health Status Family Absent Parent Involvement Family Stressful Events Parental Involvement Adoption Family Structure (1 parent) Parental Monitoring Age of Mother at Birth Family Support Parental School Involvement Assets/Income Family Trauma Parental Child Attachment Basic Needs Home Educational Resou r ces Parent Child Conflict Breast Feeding Homelessness Parenting Style Child Support Life Stressors Perceived Safety Counseling Maltreatment/Abuse Poverty or Economic Status Cultural Factors Maternal Depression Relationships and Dating Delinquency Maternal Employment Religion Family Employment Student Maternal Health Religiosity Religion English Fluency ESL Number of Siblings Residential Father Ethnic Identity Social Support Family Cohesion Parental Academic Expect. Supervision of Homework Family Management Parental Distress Supportive Adults Family Dropped Out Parental Education Television and Computer Use Peer Bullying Peer Acceptance Peer Religion Peer Academic Performance Peer Behaviors Peer Support School Academic Engagement Grade Retention School Relevance Academic Expectations Homework School Safety Academic Self Efficacy Music Instruction School SES Achievement Gap School Behaviors School Size Attendance School Belonging School/Residential Mobility Class Size School District Size Supportive School Environ. Educational Support School Minority Rates Teacher Relationships Family Type School School Quality Teacher Support Neighborhood After school Programs Immigrant % Community Neigh. Resources Crime (Neighborhood) Mentors Neigh. Violence Daycare and Preschool Neighborhood Characteristics Neigh. Youth Behaviors Extra Curricular Activities Neighborhood Deterioration Urbanicity (Rural/Urban/City) Foster Care or Public Care Neighborhood Quality

PAGE 143

133 Appendix A (Continued) Figure 3 Initial Ecological Model of School Related Factors

PAGE 144

134 Appendix A (Continued) Table 5. Demographic Characteristics of Sample Total (N = 217) School A (N = 200) School B (N = 10) School C (N = 10) Grade 09 42 (19.4%) 42 (21.3%) 0 (0.0%) 0 (0.0%) 10 51 (23.5%) 49 (24.9%) 1 (10.0%) 1 (10.0%) 11 67 (30.9%) 58 (29.4%) 2 (20.0%) 7 (70.0%) 12 57 (26.3%) 48 (24.4%) 7 (70.0%) 2 (20.0%) Gender Male 44 (20.3%) 38 (19.3%) 6 (60.0%) 0 (0.0%) Female 173 (79.7%) 159 (80.7%) 4 (40.0%) 10 (100.0%) Race White 135 (62.2%) 126 (64.0%) 8 (80.0%) 1 (10.0%) Hispanic 51 (23.5%) 50 (25.4%) 0 (0.0%) 1 (10.0%) Black 20 (9.2%) 11 (5.6%) 1 (10.0%) 8 (80.0%) Asian 4 (1.8%) 4 (2.0%) 0 (0.0%) 0 (0.0%) Multi Racial 6 (2.8%) 5 (2.5%) 1 (10.0%) 0 (0.0%) American Indian 1 (0.5%) 1 (0.5%) 0 (0.0%) 0 (0.0%) Free/Reduce Lunch No 134 (61.8%) 127 (64.5%) 7 (70.0%) 0 (0.00%) Yes 83 (38.2%) 70 (35.5%) 3 (30.0%) 10 (100.0%) Living Situation Live with 1 adult 47 (21.7%) 42 (21.3%) 3 (30.0%) 2 (20.0%) Live with 2 adults 146 (67.3%) 135 (68.5%) 6 (60.0%) 5 (50.0%) Live alone 6 (2.8%) 2 (1.0%) 1 (10.0%) 3 (30.0%) Another situation 18 (8.3%) 18 (9.1%) 0 (0.0%) 0 (0.0%) Mean Age 17.00 (SD = 1.22) 16.87 (SD = 1.18) 18.08 (SD = .90) 18.53 (SD = .49) Sample School Size 2,212 1,932 231 49 Letters Home 1,974 1,808 134 32 Response Rate 10.99% 10.89% 7.46% 31.25%

PAGE 145

135 Appendix A (Continued) Table 6. Descriptive Information for Study Factors Mean Standard Deviation Range Academic Engagement 6.17 1.62 3 9 Academic Expectations 14.47 2.49 6 17 Academic Self Efficacy 26.46 5.08 7 35 Attendance (% Present) 88.90 11.75 39.88 99.42 Educational Support 31.04 6.47 14 42 Grades Repeated 1.18 .46 1 4 Home Academic Environ. 17.68 4.024 8 24 Needs Scale .51 1.06 0 7 Parent Educational Support 13.36 3.10 6 18 School Behaviors 15.59 3.77 11 30 School Belonging 9.75 1.66 5 12 School Mobility 1.59 .97 1 4 School Relevance .97 .18 0 1 School Safety 21.86 5.23 11 33 Supportive School Environ. 19.55 3.60 7 28 Teacher Support 23.53 4.42 8 32

PAGE 146

136 Appendix A (Continued) Table 7. Sample versus Population for Demographic and Control Information Sample (N = 217) Population (N = 1995) Total (N = 2212) Significance Cumulative GPA t (274.1) = 3.37 p < .001 Mean (SD) 2.85 (.77) 2.58 (.90) 2.60 (.89) Range 0 4 0 4 0 4 Free Lunch 2 (1, N = 2212) = .307, p = .579 No 134 ( 61.8% ) 1270 ( 63.7% ) 1404 ( 63.5% ) Yes 83 (36.3%) 725 (36.3%) 808 (36.5%) Gender 2 (1, N = 2212) = 78.73 p < .001 Male 44 (20.3%) 1037 (52.0%) 1058 (48.9%) Female 173 (79.7%) 958 (48.0%) 1131 (59.1%) Grade 2 (3, N = 2212) = 7.593, p = .055 9 th 42 ( 19.4% ) 532 ( 26.7% ) 574 ( 25.9% ) 10 th 51 ( 23.5% ) 508 ( 25.5% ) 559 ( 25.3% ) 11 th 67 ( 30.9% ) 547 ( 27.4% ) 614 ( 27.8% ) 12 th 57 ( 26.3% ) 408 ( 20.5% ) 465 ( 21.0% ) Race 2 ( 4 N =2212 ) = 32.32, p < .001 White 135 (62.2%) 1347 (67.5%) 1482 (67.0%) Black 20 (9.2%) 321 (16.1%) 341 (15.4%) Hispanic 51 (23.5%) 219 (11.0%) 270 (12.2%) Asian 4 (1.8%) 46 (2.3%) 50 (2.8%) American Indian 1 (0.5%) 6 (0.3%) 7 (0.3%)

PAGE 147

137 Appendix A (Continued) 12 1.00 ** significant at the .01 level (2 tailed) significant at the .05 level (2 tailed) Table 8. Correlations Among Risk and Promotive Factors 11 1.00 .136 ** 10 1.00 .267 ** 247 ** 9 1.00 .034 .009 .104 8 1.00 .137 ** .013 .041 321 ** 7 1.00 .034 .099 .311 ** .152 .240 ** 6 1.00 .184 ** .142 .185 ** .126 .281 ** .124 5 1.00 .238 ** .212 ** .307 ** .153 .276 ** .087 .254 ** 4 1.00 198 ** .393 ** .398 ** .075 .156 .413 ** .363 ** .305 ** 3 1.00 .415 ** .389 ** .315 ** .206 ** .412 ** .328 ** .231 ** .055 .293 ** 2 1.00 .379 ** .188 ** .373 ** .307 ** .164 ** 244 ** .242 ** .254 ** .087 .308 ** 1 1.00 .375 ** .606 ** .463 ** .533 ** .281 ** .280 ** .399 ** .281 ** .364 ** .157 ** .353 ** 1) Cumulative GPA 2 ) Poverty 3) Academic Expectations 4) Academic Self Efficacy 5) Attendance 6) Educational Support 7) Homework 8) Grade Retention 9) Music Playing 10) School Behaviors 11) School Belonging 12) School Mobility

PAGE 148

138 Appendix A (Continued) Table 9. Unique Contribution of CRI Factors Regression Model B SE B VIF Step 1 (Constant) 8.156 1.177 Gender .246 .117 .128 1.005 SES (log) 3.210 .708 .328 1.422 Black ( vs White ) .563 .177 .211 1.203 Hispanic ( vs White ) .052 .130 .028 1.397 Asian ( vs White ) .576 .352 .101 1.026 Multi racial ( vs White ) .380 .291 .081 1.043 American Indian ( vs White ) .772 .695 .068 1.011 Step 2 (Constant) .343 1.104 Gender .148 .088 .077 1.014 SES(log) .428 .572 .044 1.625 Black ( v. White ) .250 .140 .094 1.208 Hispanic ( vs White ) .064 .099 .035 1.403 Asian ( vs White ) .488 .262 .085 1.026 Multi racial ( vs White ) .513 .212 .109 1.045 American Indian ( vs White ) .641 .508 .056 1.016 Percent Days Present .016 .003 .247 ** 1.273 Academic Self Efficacy .031 .009 .202 ** 1.454 Academic Expectations .091 .019 .293 ** 2.016 Grades Repeated .274 .085 .164 ** 1.267 School Behavior .163 .047 .178 ** 1.389 Significant at p < .05, ** Significant at p < .01 Note: R 2 = 229 R 2 = .394 for Step 2 ( p < .001)

PAGE 149

139 Appendix A (Continued) Table 10. Unique Contribution of ARI Factors Regression Model B SE B VIF Step 1 (Constant) 8.156 1.177 Gender .246 .117 .128 1.005 SES (log) 3.210 .708 .328 * 1.422 Black ( vs White ) .563 .177 .211 1.203 Hispanic ( vs White ) .052 .130 .028 1.397 Asian ( vs White ) .576 .352 .101 1.026 Multi racial ( vs White ) .380 .291 .081 1.043 American Indian ( vs White ) .772 .695 .068 1.011 Step 2 (Constant) .229 1.053 1.095 Gender .156 .086 .081 1.659 SES(log) .397 .539 .041 1.433 Black ( vs White ) .239 .136 .090 1.603 Hispanic ( vs White ) .097 .098 .053 1.067 Asian ( vs White ) .477 .253 .083 1.048 Multi racial ( vs White ) .510 .206 .108 1.025 American Indian ( vs White ) .623 .493 .055 1.419 Percent Days Present .016 .003 .249 ** 1.556 Academic Self Efficacy .029 .008 .191 ** 1.278 Academic Expectations .278 .081 .166 ** 1.400 Grades Repeated .172 .046 .189 ** 1.967 School Behavior .085 .019 .272 ** 1.210 Music Playing .140 .073 .090 1.095 Significant at p < .05, ** Significant at p < .01 Note: R 2 = 229 R 2 = .399 for Step 2 ( p < .001)

PAGE 150

140 Appendix A (Continued) Table 11. Descriptive Information for Dependent and Independent Factors Cumulative GPA SES Transformed SES CRI ARI Mean (SD) 2.85 (.77) 0.00 (.81) 1.69 (.08) 1. 17 (1.28) 0.06 (1.95) Range 0 4 0.63 3.71 1.62 1.98 0 5 4 5 Variance 0.60 0.65 0.01 1.65 3.82 Skewness 0.74 1.52 0.98 0.92 .16 Kurtosis 0.02 3.14 0.52 0.05 0.50

PAGE 151

141 Appendix A (Continued) Table 12. ARI Risk, Non Risk, and Promotive Factors Total Risk Non Risk Promotive Academic Expectations Range 6 17 6 14 15 16 17 n (%) 217 (100.0) 73 (33.6) 107 (49.3) 37 (17.1) Mean Score (SD) 14.47 (2.49) 11.55 (2.11) 15.58 (.50) 17 (0) Mean GPA (SD) 2.85 (.77) 2.34 (.81) 3.02 (.62) 3.38 (.50) Academic Self Efficacy Range 7 35 7 23 24 30 31 35 n (%) 217 (100.0) 63 (29.0) 105 (48.4) 49 (22.6) Mean Score (SD) 26.46 (5.08) 20.19 (2.89) 27.18 (1.88) 32.98 (1.30) Mean GPA (SD) 2.85 (.77) 2.39 (.78) 2.91 (.67) 3.32 (.66) Attendance (Risk Only) Range 39 100 0 79.99 80 100 n (%) 217 (100.0) 32 (14.7) 185 (85.3) Mean Score (SD) 88.90 (11.75) 64.40 (12.88) 92.97 (4.64) Mean GPA (SD) 2.85 (.77) 1.97 (.73) 3.01 (.67) Educational Support Range 14 42 14 26 27 36 37 42 n (%) 217 (100.0) 54 (24.9) 116 (53.4) 47 (21.7) Mean Score (SD) 31.04 (6.47) 22.31 (3.43) 31.76 (2.85) 39.30 (1.46) Mean GPA (SD) 2.85 (.77) 2.50 (.77) 2.89 (.70) 3.18 (.80) Grade Retention (Risk Only) Range 1 4 2 4 1 n (%) 217 (100) 33 (15.2) 184 (84.8) Mean Score (SD) 1.18 (.46) 2.18 (.47) 1 (0.00) Mean GPA (SD) 2.85 (.77) 2.14 (.77) 2.98 (.70) Homework (Risk Only) Range 1 5 1 2 5 n (%) 217 (100) 22 (10.1) 195 (89.9) Mean Score (SD) 3.50 (1.34) 1 (0) 3.78 (1.10) Mean GPA (SD) 2.85 (.77) 2.38 (.90) 2.91 (.74) Music Playing (Promotive Only) Range 0 1 0 1 n (%) 217 (100) 118 (54.4) 99 (45.6) Mean Score (SD) .46 (.50) 0 (0.00) 1 (0.00) Mean GPA (SD) 2.85 (.77) 2.66 (.80) 3.09 (.067) School Behaviors (Risk Only) Range .87 4.14 (100) .33 4.14 .80 .32 n (%) 217 (100) 52 (32.3) 165 (67.7) Mean Score (SD) 0.00 (.79) .38 (.31) 1.19 (.90) Mean GPA (SD) 2.85 (.77) 2.21 (.66) 3.06 (.72) School Belonging (Promotive Only) Range 5 12 5 11 12 n (%) 217 (100) 180 (82.9) 37 (17.1) Mean Score (SD) 9.75 (1.66) 9.29 (1.44) 12 (0.00) Mean GPA (SD) 2.85 (.77) 2.78 (.74) 3.02 (.84) School Mobility (Risk Only) Range 0 3+ 2 + 0 1 n (%) 217 (100) 36 (16.6) 181 (83.4) Mean Score (SD) 0.59 ( 2.56 (.504) 0.19 (.40) Mean GPA (SD) 2.85 (.77) 2.30 (.72) 2.96 (.74)

PAGE 152

142 Appendix A (Continued) Table 13. CRI Risk Factors Total Risk Non Risk Academic Expectations Range 6 17 6 14 15 17 n (%) 217 (100.0) 73 (33.6) 144 (66.4 ) Mean Score (SD) 14.47 (2.49) 11.55 (2.11) 15.94 (.76) Mean GPA (SD) 2.85 (.77) 2.34 (.81) 3.11 (.61) Academic Self Efficacy Range 7 35 7 23 24 35 n (%) 217 (100.0) 63 (29.0) 154 (71.0) Mean Score (SD) 26.46 (5.08) 20.19 (2.89) 29.03 (3.20) Mean GPA (SD) 2.85 (.77) 2.39 (.78) 3.04 (.69) Attendance Range 39 100 0 79.99 80 100 n (%) 217 (100.0) 32 (14.7) 185 (85.3) Mean Score (SD) 88.90 (11.75) 64.40 (12.88) 92.97 (4.64) Mean GPA (SD) 2.85 (.77) 1.97 (.73) 3.01 (.67) Educational Support Range 14 42 14 26 27 42 n (%) 217 (100.0) 54 (24.9) 163 (75.1) Mean Score (SD) 31.04 (6.47) 22.31 (3.43) 33.93 (4.25) Mean GPA (SD) 2.85 (.77) 2.50 (.77) 2.97 (.74) Grade Retention Range 1 4 2 4 1 n (%) 217 (100) 33 (15.2) 184 (84.8) Mean Score (SD) 1.18 (.46) 2.18 (.47) 1 (0 .00 ) Mean GPA (SD) 2.85 (.77) 2.14 (.77) 2.98 (.70) Homework Range 1 5 1 2 5 n (%) 217 (100) 22 (10.1) 195 (89.9) Mean Score (SD) 3.50 (1.34) 1 (0 .00 ) 3.78 (1.10) Mean GPA (SD) 2.85 (.77) 2.38 (.90) 2.91 (.74) School Behaviors Range .87 4.14 (100) .33 4.14 .80 .32 n (%) 217 (100) 52 (32.3) 165 (67.7) Mean Score (SD) 0.00 (.79 ) .38 (.31 ) 1.19 (.90 ) Mean GPA (SD) 2.85 (.77) 2.21 (.66 ) 3.06 (.72) School Mobility Range 0 3+ 2 + 0 1 n (%) 217 (100) 36 (16.6) 181 (83.4) Mean Score (SD) 0.59 ( .97) 2.56 (.504) 0.19 (.40) Mean GPA (SD) 2.85 (.77) 2.30 (.72) 2.96 (.74)

PAGE 153

143 Appendix A (Continued) Figure 4. Scatter plot of Residuals for CRI Model

PAGE 154

144 Appendix A (Continued) Figure 5. Linearity of Residuals for CRI Model

PAGE 155

145 Appendix A (Continued) Figure 6. Distribution of Residuals For CRI Model

PAGE 156

146 Appendix A (Continued) Table 14. Correlation of Regression Factors for the CRI 1 2 3 4 5 6 7 8 9 1 Cumulative GPA 1.000 2 Gender .121 1.000 3 American Indian ( vs White ) .071 .034 1.000 4 Black ( vs White ) .282 ** .002 .022 1.000 5 Hispanic ( vs White ) .122 .009 .038 .177 1.000 6 Asian ( vs White ) .097 .016 .009 .044 .076 1.000 7 Multi Racial ( vs White ) .096 .055 .011 .054 .093 .023 1.000 8 SES (log) .390 ** .021 .042 .237 ** .409 ** .047 .062 1.000 9 CRI .707 ** .051 .044 .220 ** .191 ** .062 .022 .405 ** 1.000 Significant at p < 05 ** Significant at p < .01

PAGE 157

147 Appendix A (Continued) Table 15. Standard Multiple Regression Results for the CRI B SE B VIF Step 1 (Constant) 8.16 1.18 Gender 0.25 0.12 0.13 1.01 American Indian ( vs White ) 0.77 0.70 0.07 1.01 Black ( vs White ) 0.56 0.18 0.21 1.20 Hispanic ( vs White ) 0.05 0.13 0.03 1.40 Asian ( vs White ) 0.58 0.35 0.10 1.03 Multi racial ( vs White ) 0.38 0.29 0.08 1.04 SES (log) 3.21 0.71 0.32 ** 1.42 Step 2 (Constant) 5.09 0.92 Gender 0.17 0.09 0.09 1.01 American Indian ( vs White ) 0.48 0.52 0.04 1.01 Black ( vs White ) 0.29 0.14 0.11 1.24 Hispanic ( vs White ) 0.05 0.10 0.03 1.41 Asian ( vs White ) 0.80 0.27 0.14 1.03 Multi racial ( vs White ) 0.46 0.22 0.10 1.04 SES(log) 1.13 0.56 0.12 1.56 CRI 0.39 0.03 0.65 ** 1.24 Significant at p < .05, ** Significant at p < .01 Note: R 2 = .2 3 R 2 = .34 for Step 2 ( p < .001)

PAGE 158

148 Appendix A (Continued) Figure 7. Cumulative GPA by CRI Scores

PAGE 159

149 Appendix A (Continued) Figure 8. Scatter plot of Residuals for ARI Model

PAGE 160

150 Appendix A (Continued) Figure 9. Linearity of Residuals for ARI Model

PAGE 161

151 Appendix A (Continued) Figure 10. Distribution of Residuals for ARI Model

PAGE 162

152 Appendix A (Continued) Table 16. Correlation of Regression Factors for ARI Model 1 2 3 4 5 6 7 8 9 1 Cumulative GPA 1.000 2 Gender .121 1.000 3 American Indian ( vs White ) .071 .034 1.000 4 Black ( vs White ) .282 .002 .022 1.000 5 Hispanic ( vs White ) .122 .009 .038 .177 1.000 6 Asian ( vs White ) .097 .016 .009 .044 .076 1.000 7 Multi Racial ( vs White ) .096 .055 .011 .054 .093 .023 1.000 8 SES (log) .390 ** .021 .042 .237 ** .409 ** .047 .062 1.000 9 ARI .714 ** .096 .068 .195 .206 ** .013 .009 .408 ** 1.000 Significant at p < .05 ** Significant at p < .01

PAGE 163

1 53 Appendix A (Continued) Table 17. Standard Multiple Regression Results for ARI Model B SE B VIF Step 1 (Constant) 8.156 1.177 Gender .246 .117 .128* 1.005 American Indian ( vs White ) .772 .695 .068 1.011 Black ( vs White ) .563 .177 .211* 1.203 Hispanic ( vs White ) .052 .130 .028 1.397 Asian ( vs White ) .576 .352 .101 1.026 Multi racial ( vs White ) .380 .291 .081 1.043 SES (log) 3.210 .708 .328** 1.422 Step 2 (Constant) 4.65 0.93 Gender 0.11 0.09 0.06 1.01 American Indian ( vs White ) 0.30 0.53 0.03 1.02 Black ( vs White ) 0.33 0.14 0.13 1.23 Hispanic ( vs White ) 0.06 0.10 0.03 1.41 Asian ( vs White ) 0.61 0.27 0.11 1.03 Multi racial ( vs White ) 0.38 0.22 0.08 1.04 SES(log) 1.10 0.56 0.11 1.56 ARI 0.26 0.02 0.65 ** 1.25 Significant at p < .05, ** Significant at p < .01 Note: R 2 = .2 3 R 2 = .33 for Step 2 ( p < .001)

PAGE 164

154 Appendix A (Continued) Figure 11. Cumulative GPA by ARI scores

PAGE 165

155 Appendix A (Continued) Table 18. Logistical Regression Results for CRI and ARI models. 95% Confidence Interval for Odds Ratio B SE Wald 2 Test Sig. Odds Ratio Lower Upper Model 1 SES (log) 5.19 3.47 2.24 0.13 0.01 0.00 4.96 Gender 0.06 0.60 0.01 0.92 1.06 0.33 3.43 Race 3.69 0.60 Black v s White 0.82 1.23 0.44 0.51 2.27 0.20 25.14 Hispanic v s White 0.10 1.30 0.01 0.94 0.90 0.07 11.54 Asian v s White 1.16 1.27 0.84 0.36 3.19 0.27 38.24 American Indian v s White 20.97 18175.89 0.00 1.00 1274713838.05 0.00 SES (log) 20.68 40192.97 0.00 1.00 954574198.15 0.00 CRI 1.26 0.23 30.27 0.00 0.28 0.18 0.44 Constant 12.02 6.00 4.02 0.05 165815.59 Note R 2 = .30 (Cox & Snell), .51 (Nagelkerke). Model 2 (8) = 76.15 p < .001 Significant at p < .05, ** Significant at p < .01 Model 2 SES (log) 5.67 3.20 3.14 0.08 0.00 0.00 1.83 Gender 0.06 0.55 0.01 0.92 1.06 0.36 3.13 Race 4.76 0.45 Black v s White 0.48 1.20 0.16 0.69 1.62 0.15 17.07 Hispanic v s White 0.58 1.27 0.21 0.65 0.56 0.05 6.78 Asian v s White 0.88 1.24 0.51 0.48 2.42 0.21 27.43 American Indian v s White 19.83 18945.21 0.00 1.00 409142889.39 0.00 ARI 0.81 0.16 25.58 0.00 2.25 1.64 3.08 Constant 11.69 5.59 4.37 0.04 ** 119733.74 Note R 2 = .2 6 (Cox & Snell), .4 4 (Nagelkerke). Model 2 (8) = 64.72 p < .001 Significant at p < .05, ** Significant at p < .01

PAGE 166

156 Appendix A (Continued) Table 19. Area Under the Curve and Accuracy Indices for CRI and ARI Area Under the Curve SE Sig. Sensitivity Specificity 1 Specificity CRI .88 .03 .000 Cutoff 1 .971 .491 .503 Cutoff 2 .824 .743 .257 Cutoff 3 .676 .913 .087 Cutoff 4 .265 .995 .005 ARI .8 5 .04 .000 Cutoff 0 .912 .486 .514 Cutoff 1 .794 .672 .328 Cutoff 2 .735 842 .158 Cutoff 3 .471 .951 .049

PAGE 167

157 Appendix A (Continued) Figure 12. ROC Curve for the CRI

PAGE 168

158 Appendix A (Continued) Figure 13. ROC Curve for the ARI

PAGE 169

159 Figure 14. Final Model of School Factors and Cumulative GPA

PAGE 170

160 Figure 15. Final Ecological Model

PAGE 171

161 Appendix B : Concept Mapping Form

PAGE 172

162 Appendix C : Survey Instrument

PAGE 173

163 Appendix C (Continued)

PAGE 174

164 Appendix C : (Continued)

PAGE 175

165 Appendix D : Consent and Assent Forms

PAGE 176

166 Appendix D (Continued)

PAGE 177

167 Appendix D (Continued)

PAGE 178

168 Appendix D (Continued)

PAGE 179

169 Appendix D (Continued)

PAGE 180

170 Appendix D (Continued)

PAGE 181

171 Appendix E : Institutional Approval Letter

PAGE 182

About the Author a practicing school social worker in Pinellas County for the last 11 years, working with pregnant and parenting teens. Mr. Lucio is also a Licensed Clinical Social Worker in the state of Florida. maternal and child health, and technology in social work. He has published and presented at state and national school social work conferences. Mr. Luc io has also been active in the Florida Association of School Social Workers, as a board member and president.