GROUP RANDOMIZED PREVENTION TRIALS IN SCHOOLS: DESIGN AND METHODOLOGICAL CHALLENGES. Stephanie Jones1, Joshua Brown2, J. Lawrence Aber2, 1Fordham University, New York, NY United States; 2New York University, New York, NY United States
This presentation will address several critical design and methodological challenges associated with embedding basic developmental research questions in a group randomized evaluation of a school-based preventive intervention in social and emotional learning and literacy development. These challenges include: (1) employing lessons learned from previous non-experimental work to develop an understanding of and measurement strategies for the change processes expected to underlie any intervention effects (Aber et al., 1998; Aber et al., 2003); (2) understanding expected effect sizes for the sample and context being studied and recruiting enough sites to have enough power at the level of random assignment to detect these effects (Raudenbush & Xiao-Feng, 2001); and (3) designing and conducting the appropriate analyses to adequately answer these questions for school-based designs, e.g., nested longitudinal trajectory models; and cross classified models that allow tracking of students across years in the context of changing teachers. In addition, the implications of experimental evaluations of multi-level interventions for program improvement and policymaking will be discussed.
As one of seven sites involved in a national multi-site evaluation of social and character education programs funded by the Institute for Education Sciences of the Department of Education, the study of the 4Rs Program is unique. First, this study rigorously tests a program and developmental model that is based on integrating a social and emotional learning intervention within a literacy development curriculum. The program and developmental models incorporate key elements from theories of (a) children´s social-information processing (e.g., hostile attributional processing; Dodge, et al., 2002); (b) the stages of children´s development of social understanding (e.g., Selman, 2003); and (c) children´s literacy development in the elementary school years (e.g., Snow, Burns & Griffin, 1998). Second, in addition to testing intervention effects on trajectories of children´s social, emotional, and academic development, this study also examines direct intervention effects on changes in teacher´s own social and emotional skills, professional development, and provision of extended opportunities to engage youth in social and emotional learning. Third, this study tests whether and how (a) changes in teacher development and extended opportunities for social-emotional learning mediate the effect of 4Rs on children´s social-emotional and academic development, and (b) the degree to changes in children´s social-emotional development and changes in children´ academic achievement influence each other in a mediated uni-directional, or in a bi-directional manner.
NEW EMPIRICAL BAYESIAN METHODS FOR STUDYING RECIPROCAL INFLUENCE IN DYADS AS A MEDIATOR OF PREVENTIVE INTERVENTION EFFECT. Getachew Dagne1, George Howe2, 1University of South Florida, Tampa, FL United States; 2George Washington University, Washington, DC United States
Developmental researchers have long contended that reciprocal influences in families contribute to risk for behavior problems, substance use, and emotional disorders. As a result, family-based interventions often target family interaction, assuming that changes in reciprocal influence patterns in the family will in turn reduce risk for future problems. However, quantitative methods for characterizing reciprocal influence in family interaction have been lacking, and very few studies have attempted to test whether such changes in fact mediate the effects of family-based interventions on distal outcomes.
This paper presents an extension of recent work on empirical Bayesian estimation techniques for characterizing patterns of interaction based on direct observation of family interactions. Using a multilevel modeling framework, this extension allows for the characterization of reciprocal influence within individual family dyads, as well as estimation of variation in influence parameters across a sample of dyads. It also allows for modeling this variation within a regression framework, such that variation in strength of reciprocal influence can be tested as a mediator of the effects of intervention on distal outcome. An example of this model is presented, using observational data from a sample of marital dyads to characterize variation in negative reciprocity cycles, and to test whether this variation mediates the effects of recent life changes on future depression.
BAYESIAN HIERARCHICAL MULTISTATE COMPETING RISKS MODELS FOR DURATION DATA: AN APPLICATION TO A STUDY OF EMOTION REGULATION DURING FAMILY INTERACTION. Getachew Dagne1, 1University of South Florida, Tampa, FL United States, Getachew A. Dagne, Mike Stoolmiller, and James Synder
We propose a Bayesian continuous-time Semi-Markov model for multilevel duration data. The model is developed for the analysis of repeating-events, competing-risk behavioral observation data where there are multiple origin states and multiple types of event (competing risks). The different transitions are modeled jointly to allow for correlation across transitions. The correlation issue is addressed by the use of random effects for each transition hazard rate. The random effects for the different transition rates may be correlated within dyads, but are assumed to be independent for different dyads. The model is applied in an analysis of the School Transitions Project (“OZ”) - a longitudinal study of children's behavioral and Social adjustment at the transition to elementary school through 3rd grade, and the ascertainment of parent-child and peer social relational processes that affect that adjustment. The covariates used in the modeling process are marital transitions, child conduct problems, child depression, and parental depressive symptoms.
POSTERS: MIDDLE CHILDHOOD DEVELOPMENT
SOCIAL INFORMATION PROCESSING MODEL : VALIDATION OF AN INSTRUMENT THAT INTEGRATES EMOTIONAL PERCEPTIONS IN SIMULATED SITUATIONS OF PROVOCATION. Normand Rondeau1, François Bowen1, Michel Fournier2, 1Universite de Montreal, Montreal, Quebec Canada; 2Direction de la santé publique - Montréal Centre, Montréal, Quebec Canada
The Social Information Processing model – SIP - (Crick & Dodge, 1994) has been used mostly to understand externalized disorders such as reactive and proactive aggression, and social rejection. Lemerise & Arsenio (2000) propose modifications to the original model allowing integration of the emotional dimension. Inspired of their work, we are testing a hypothesis which consists of integrating the child´s perception of his own affective state, that of the peer and the recognition of emotions will have an influence on the steps of the SIP.
Based on the “Social problem solving interview” – SPSI – (Bream, Hymel & Rubin, 1986; Rubin, Bream & Rose-Krasnor, 1991), we have slightly modified the interview outline so as to include questions that will permit a better understanding of the child´s encoding and interpretation of a situation of provocation. Also, instead of using hypothetical role-playing where the actors´ intentions are ambiguous, the role-playing used involves a clearly conflictual issue where the relationship/situation is threatened. The modified SPSI was collected from 1673 children between the ages of 5 and 12 years (51.2 % girls).
A series of analyses (correspondence analysis, cluster analysis, binary decision tree analysis) lead to results where the children´s responses are grouped into 11 profiles of social information processing. From the 11 profiles, emotion (the child´s or the peer´s) is among the main descriptors in three profiles (5, 7, 10). However, it appears in almost each profile with a variable relative weight. The main descriptors (from up to 25) found in these profiles are the following: 1) Do not know (7.7%); 2) Do not know and authority help-seeking (13.5%); 3) Relationship and situation restored ( 4.9%); 4) Relationship altered and the problem is the peer (6.1%); 5) Passive avoidance and neutral emotion (5.0%); 6) Deteriorated relationship and situation and aggressive solution (10.5%); 7) Inference reaction of peer and peer is upset (6.2%); 8) Prosocial solution and the situation is the problem (9.2%); 9) Relationship and situation restored and the problem is solved (23.6%); 10) Child and peer feel bad (7.3%); 11) Evaluation of the solution as satisfying (6.0%).
Emotion and other steps of the SIP are combined to form profiles. The next stage will be to describe profile in conjunction with behavioral characteristics. Taking account of the nature of profiles, the discussion will examine implications for intervention on problem behaviors.
TRY OUT AND EVALUATION OF THE GOOD BEHAVIOR GAME IN BELGIUM. Hilde Colpin1, Sofie Driesmans1, Lien Willem1, 1Catholic University of Leuven, Tiensestraat 102, 3000 Leuven Belgium
The Good Behavior Game is a universal school-based prevention program designed to improve elementary school children´s social adaptation to the classroom related to rules and authority. Randomized control studies in the US and the Netherlands have found significant reductions in teacher and peer ratings of aggression as well as teacher ratings of shy behaviour (a strong risk factor for negative outcomes when coupled with early aggression) and on task behaviour rated by independent observers (e.g. Kellam, Rebok, Mayer & Ialongo, 1994; Kellam, Ling, Merisca, Hendricks Brown, & Ialongo, 1998; Ialongo, Werthamer, Kellam, Brown, Wang, & Lin, 1999; van Lier, Muthén, van der Sar & Crijnen, 2004). A small-scale try out with the GBG has been set up to investigate the feasibility and effectiveness of the program in Belgium; in a second phase the program will be implemented and tested on a larger scale. In one school in the metropolitan area of Brussels (with a relatively high proportion of children from disadvantaged and allochtonous families), the four classes of the second grade were randomly assigned to either the intervention condition (two classes, n=37) or the control condition (two classes, n=38). Data-collection is conducted at baseline (Autumn 2004) and immediately after the end of the intervention (Spring 2005) and consists of teacher ratings of children´s problem behavior (PBSI-R, van Lier et al., 2004), peer nominations of children´s problem behavior (Dolan, Kellam, Hendricks Brown, Werthamer-Larsson, Rebok, Mayer et al., 1993) and acceptance by peers (Coie, Dodge & Coppotelli, 1982) and an assessment of children´s self-representations (Pictorial Self-Evaluation Scale, Verschueren, Marcoen & Schoefs). The quality of program implementation is assessed by observations in the classroom. Results of this evaluation study will be presented at the conference.
PREVENTING SCHOOL BULLYING BY USING ENVIRONMENTAL "HOTSPOTS.". Lisa Rapp-Paglicci1, Catherine Dulmus2, Karen Sowers2, William Rowe3, 1University of South Florida, Lakeland, FL United States; 2University of Tennessee, Knoxville, TN United States; 3University of South Florida, Tampa, FL United States
Bullying is a chronic and ubiquitous problem that has ignited concern about children´s safety at school. Most studies completed thus far, have focused on the interpersonal psychological characteristics of bully and victim, but little has been examined regarding common locations or hotspots for bullying. This study found location to be of import, in that, hotspots can be identified at each school and differ with regard to gender and grade. This study utilized a survey research design to gather information from students that included questions related to perceptions and observations of types of bullying among students and locations in the school setting where bullying behaviors take place. The convenience sample consisted of 192 students in grades three through eight. Subjects were recruited from three rural public schools (an elementary school, a middle school, and a school that housed grades K-8) located within the same school district in a Southeastern rural region of the United States. Results indicated more elementary school students being bullied on the playground than middle school students, who were most often bullied in classrooms and hallways. Differences between gender and location indicate that males are most often bullied on the playground, while females are most often bullied in classrooms. This data supports the premise that a comprehensive approach including interpersonal interventions as well as specific environmental modifications would be better served. Schools interested in curbing bullying need to assess the location of bullying particular to their own property, and address those areas specifically. Schools can prevent and intercede more effectively with bullying if they can identify their particular hotspots and provide a comprehensive intervention that includes targeting specific locations, educating school personnel, as well as providing interpersonal treatment to bully and victim. Our full development of this area has yet to be achieved. Research is critically needed, in order to further understand hotspots, their relation to types of bullying behaviors and their implication for prevention of bullying.
TEMPERAMENT PATTERNS FROM AGE 18 MONTHS TO 8.5 YEARS IN A LONGITUDINAL STUDY: DEVELOPMENT AND ASSOCIATION WITH BEHAVIOR PROBLEMS. Harald Janson1, Kristin Mathiesen2, 1Atferdssenteret Unirand/Center for the Study of Conduct Problems & Innovative Practice, Oslo, Oslo Norway; 2Regional Center for Child and Adolescent Mental Health R.BUP, Oslo, Oslo Norway
Background: Early childhood temperament is a strong individual predictor of antisocial behavior, and as such of great interest for screening for very early problem-behavior development. Conceptualizations of temperament have most often been variable-oriented; predictions from classifications based on configurations of values have less seldom been addressed. Also, the contribution of parent-completed questionnaire reports of young children's temperament in screening for behavior-problem development needs to be investigated further.
Objectives: To find a replicable and meaningful cluster solution for describing children's temperaments profiles. To describe development from age 18 months to age 8.5 in terms of cluster membership. To explore the concurrent and predictive associations between temperament cluster membership and problem behavior.
Method: We used data from an ongoing prospective, longitudinal study of children in Oslo representative of the general population, described in more detail elsewhere. The present study analyzed mother-completed questionnaires of child temperament (Buss and Plomin's EAS) and problem behavior (Richman & Graham's BCL, or Goodman's SDQ) from ages 18 months (n=921), 2.5 years (n=784), 4.5 years (n=737), and 8.5 years (n=512). A modification of Bergman's and El-Khouri's ISOA procedure was implemented: Cluster analysis was performed independently on four weighted standardized variables of the EAS questionnaire in two half-samples of 1,024 temperament profiles. All profiles were re-classified onto the closest cluster in a final four-cluster solution. Development in terms of cluster membership was investigated by inspecting statistical types and anti-types in cross-tabulations of cluster memberships at adjacent time points.
Results: A four-cluster solution proved highly interpretable and replicable. The development in terms of cluster membership proved highly stable: Staying in the same cluster over a time interval was always a statistical type, and all antitypes represented changing cluster classifications. There were strong and consistent concurrent associations between cluster membership and problem behavior at all time points. Also, cluster membership prospectively predicted behavior problems over all time intervals studied.
Conclusion: Empirically derived and conceptually meaningful temperament types proved highly stable over time, and showed strong and consistent associations with problem behavior. These results support further investigations of the usefulness of a pattern-oriented approach to screening for problem-behavior development.
THE MEDIATING ROLE OF PARENT AFFECT IN THE RELATIONSHIP BETWEEN PARENT SUBSTANCE USE AND CHILD ADJUSTMENT. Nathaniel Riggs1, Mary Ann Pentz1, Chih-Ping Chou1, 1University of Southern California, Alhambra, CA United States
One in four youth live with a parent who abuses substances. This is an issue of substantial public health importance as there is a well-established link between parental substance use and developmental problems such as physical, academic, and social-emotional problems in their offspring. Given this growing literature linking parent substance use to youth adjustment, very little is known with respect to any potential mediating pathways through which this relationship is manifested. In this paper, we develop and test a conceptual model that outlines the relationships among parent substance use; two potential intervening factors, parent anger and depression; and child cognitive and psychological adjustment. This population-based study used structural equation models to demonstrate that parent affect is an intervening factor in the relationship between parent substance use variables and the anxiety, conduct problems, distractibility, and impulse control of their children (n = 462). Substance use variables included illicit drugs not including marijuana, marijuana, alcohol, and problems associated with alcohol. Analyses suggest that both parent anger and depression are important variables in the relationship between these parent substance use variables and offspring adjustment. One important implication is that for intervention efforts targeting youth with substance using parents to prove most effective, a focus must not only be placed on parent substance use itself, but also on parent regulation of affect.
INVOLVEMENT WITH DELINQUENT PEERS AND ALCOHOL USE. Ley Killeya-Jones1, Patrick Malone1, Megan Golonka1, Philip Costanzo1, 1Duke University, Durham, NC United States
The substantial evidence in the literature affirming the importance of peers´ substance use and beliefs as a risk factor for young adolescents´ own substance use derives in general from mostly European-American samples (Aseltine, 1995; Hawkins, Catalano, & Miller, 1992; Newcombe, Mahhadian, & Bentler, 1986; Oetting & Beauvais, 1987). Other evidence suggests that peers are not as important for understanding substance use among African-American adolescents (Giordano et al., 1986; Giordano et al., 1998). In their seminal paper on the influence of peers on adolescent delinquency, Giordano et al (1986) reported that delinquent friends have greater effects on adolescent delinquency when the friendships are higher in attachment and contact. African-American adolescents report less peer pressure and a lower need for approval from their friends, and describe their friendship relationships to be less intimate (Giordano, Cernkovich, & DeMaris, 1993). Farrell and Danish (1992) reported that changes in gateway drug use were not associated with peers models or peer pressure in their mainly African-American sample (92%).
These findings suggest that involvement with peers is likely to mediate the relationship between drug using peers and adolescents´ substance use behavior for European-American but not African-American adolescents, and it is this hypothesis that we test in the present study. The data used derive from a NIDA-supported survey-based investigation of a wide array of peer influences on substance use. For the current study, we examine the effects of adolescents´ involvement with delinquent peers, and peer drug use on alcohol use, deriving separate logistic regression models to predict current alcohol use in a middle-school sample of 131 European-American (n = 76) and African-American (n = 55) adolescents. Current alcohol use was conceptualized as a dichotomous variable, use/no use in the past 30 days.
Preliminary analyses show that African-American and European-American adolescents exhibit similar rates of lifetime, current, and binge drinking. Peer substance use was a significant predictor of lifetime and past 30 days alcohol use for both groups, but there were no significant predictors of binge drinking. Although the degree of involvement with friends was similar across groups, involvement with friends was significantly and positively related to current alcohol use only for European-American adolescents, in support of our hypothesis. However, the interaction between friends´ use and involvement did not reach significance for either group. Further planned analyses with an increased sample size will investigate these relationships in more detail and will inform the results presented.
RELATIONAL AGGRESSION IN ADOLESCENT GIRLS: LINKS TO SOCIAL INFLUENCE AND RISK FACTORS FOR ALCOHOL USE. Christina Grimes1, Martha Putallaz1, Shari Miller-Johnson1, Ley Killeya-Jones1, Kristen Foster1, Megan Golonka1, 1Duke University, Durham, NC United States
Relational aggression in children has been linked to peer rejection and adjustment difficulties. However, for adolescent girls, it is related to high clique centrality and perceived popularity (but not sociometric popularity), suggesting that relational aggressors wield considerable social power in the peer culture. Whether they are at risk for adjustment difficulties is unknown. The current study is an initial exploration of relational aggression´s link to risk factors in one adjustment domain, that of alcohol use, and whether its social power extends to peer influence in that domain.
Participants were 150 7th-graders (89 girls; 51% European American, 37% African American, 12% other ethnicities) involved in a comprehensive study of peer influence processes affecting substance use beliefs and behaviors. The findings reported here include data on only the 89 girls in the sample, but all 150 students completed sociometric surveys, social cognitive maps (SCM), and self-report measures. For the sociometric measure, students were provided with a roster of all 7th graders in their school and were asked to make unlimited nominations of peers who fit various descriptors, including perceived popularity, overt and relational aggression, and three types of leadership: conventional (“leaders and good to have in charge”); trendsetting (“other kids listen to; set the trends”); and deviant (“good at getting other kids to break the rules”). The SCM measure asks students to list all of the groups they can think of in their grade, and assesses clique structure and each student´s centrality in the social network. Self-report measures included items such as self-ratings on personal qualities, substance use, sensation seeking, susceptibility to peer pressure, and ability to influence others.
Peer nominations for overt and relational aggression were highly correlated (r = .65, p < .001). With overt aggression controlled for in all analyses, relational aggression was significantly correlated with perceived popularity, social network centrality, and trendsetter leadership (r´s = .54, .54, and .55 respectively, all p < .001), but not with conventional or deviant leadership.
Peer-nominated relational aggression was related to generally positive views of the self, including positive correlations with self-ratings on coolness, attractiveness, having lots of friends, and having power to decide who´s in or out (r´s = .42, .33, .35, and .28, all p < .01). However, it was also correlated with sensation seeking, susceptibility to peer pressure, and higher levels of alcohol use (r´s = .29, .31, and .30, all p < .01). Follow-up surveys in October 2004 will explore the stability of these factors.