Citation

Walker HM, Stieber S, Ramsey E, O'Neill R. J. Child Fam. Stud. 1993; 2(4): 295-315.

Copyright

(Copyright © 1993, Holtzbrinck Springer Nature Publishing Group)

DOI

10.1007/BF01321227

PMID

unavailable

Abstract

This study investigated the efficacy of three different regression models in predicting arrest rate over the fifth to seventh grade range for a sample of antisocial and at risk control middle school boys (N=76). These boys were selected from a larger sample of approximately 200 boys and their families for inclusion in a long term longitudinal study beginning in the fourth grade. Subjects in the present study were identified when they were in the fifth grade and school measures of adjustment were recorded annually for them on two occasions (fall, spring) through the ninth grade. Predictors were derived from the following variables: (a) teacher ratings of social skills; (b) direct observations of academic engagement in classroom settings; (c) direct observations of the playground social behavior of target subjects and peers; and (d) discipline contacts with the principal's office as extracted from archival school records. These predictor variables appeared to assess respectively teacher related and peer related forms of school adjustment. The regression models investigated yielded R-Squares of approximately 30 between arrest rate and two to three predictor variables in each analysis. However, following the deletion of two subject cases that represented extreme instances of underprediction (i.e., subjects who were arrested [i.e., seven arrests each] but whose fifth grade behavioral profiles suggested they would not be), the obtained R-Squares increased to approximately .55. Implications of the findings for the early screening and identification of at risk students are discussed.