Multivariate summer school effects

The effect of assignment to and participation in summer school for a moderately at-risk sample of kindergarten students was examined with multivariate analytic methods. A multivariate analysis of variance applied to difference scores capturing the change in summer literacy outcomes revealed that kindergarten students randomly assigned to summer school outperformed their control group peers on a linear composite of early literacy indicators. The estimated group difference was greater when participation in summer school was distinguished from receipt of the summer program offer in analyses that explicitly adjusted for the proportion of students who failed to comply with their assignment. These results demonstrate that the nature and generalizability of the inference regarding program performance varies in relation to the intended and achieved design and the analytic model applied to data. Implications for the evaluation of summer school programs are discussed.