Effects of the Lessons in Character English Language Arts Character Education Program on Behavior and Academic Outcomes
Analysis plan
To estimate program impacts, outcomes for students and classrooms in treatment schools are compared with those for students and classrooms in control schools. Multilevel regression models are used to analyze the effects of Lessons in Character, to account for data clustering by school (Goldstein 1987; Raudenbush and Bryk 2002; Murray 1998). The impact analyses control for all baseline (pretest) measures of outcome variables and other student- and school-level covariates.
The procedures described by Schochet (2008) are used to account for testing multiple hypotheses involving the study's numerous outcome variables. Specifically, four outcome domains are delineated—student academic achievement, social skills, and problem behaviors and school and classroom climate—with primary and secondary outcomes within each domain. Multiple comparison procedures are used for each outcome domain to reduce the probability of finding statistically significant program impacts when impacts are due to chance alone. Exploratory analyses also examine potential impacts on other areas, such as staff culture of belonging and student academic engagement, cooperation, and disciplinary referrals. In addition to estimating program impacts, the study also conducts exploratory analyses to investigate differences in program impacts by gender, race/ethnicity, and limited English proficiency. Multiple comparison procedures are not used for the exploratory analyses of outcomes.