The adaptive behavior scale: a study of criterion validity.

Abstract

The validity of the Adaptive Behavior Scale (ABS) for placement purposes was estimated. Subjects were 95 formerly institutionalized retarded persons, 97 institutional residents who were referred for discharge, and 178 current institutional residents. A stratified random-sampling process was used to create two equal-size samples. The ABS domain scores of the 185 individuals in the derivation sample were submitted to stepwise multiple discriminant function analysis. A linear combination of seven ABS domains explained 34 percent of the variance in group membership and correctly classified 60 percent of the derivation sample. This same combination correctly classified 49 percent of the cross-validation sample. Factor scores were calculated and submitted to similar statistical analysis. A linear combination of three factors correctly classified 54 percent of the derivation sample, and no shrinkage was observed upon cross-validation. Results suggest that knowledge of an individual's ABS scores would enable a test user to make estimates of group membership. Additionally, factor scores, with their increased reliability and greater predictive validity upon cross-validation, appeared to be a more appropriate unit of analysis than were domain scores for the ABS.