Archival Version

The data for this study was drawn from a panel of students enrolled in the training and control schools (10 schools in each group) in the state between the 2013/14 and 2017/18 academic years. The core dataset was provided by the state's Department of Education. Based on a large longitudinal sample including an intervention and control condition, program effects were evaluated across 5 years, including 2 years before and 2 years after the training. The training, occurring in the 2015/16 academic year, serves as the "training year" for this study. School-level data was the unit of analysis in the first set of analyses. Training effects on school wide outcomes were tested using Piecewise hierarchical linear models. Next, logistic multilevel models were run separately for each year to explore the impacts of gender,race, and age on the outcomes. Students were the unit of analysis for the second set of analyses, taking into account non-independence of observations.