Multilevel modeling of behavioral observations of Head Start students; Accounting for initial conditions

Abstract

A sample of 101 Head Start students were repeatedly tested for behavioral improvements during a TCIT study. The measure used to test the students (the BOPS) had not previously been validated. Factor Analysis was used to determine the underlying structure of the BOPS; a frequency-based observation tool used to measure challenging and prosocial behavior in preschool students. Parallel analysis and a scree plot were used to select the five factors retained from the Principal Axis Factoring (KMO=.768). The five factors retained accounted for over 62% of the variance. Three of the factors were prosocial in nature and two were challenging behaviors. Multilevel Model was used to demonstrate the nested nature of the students over multiple occasions in a classroom structure. Variance components were significant (p<.001) and the ICC=80.18%, indicating that the nested model was appropriate. The time variant model was significantly better than the null (X2(2)=56.73, p<.001).