Initial tests
Regression models
Because there were differences between the control class and the two experimental classes on several predictor variables, the bivariate test results cannot be taken at face value. The significant improvement in student performance in fall 2012 and the slight improvement in student performance in spring 2013 might result from the differences in the composite ACT scores and cumulative GPAs of the students in those semesters, for example. Multivariate modeling was therefore conducted to control for the influence of several predictor variables simultaneously. Two multivariate ordinary least squares (OLS) regression models were constructed, one for the initial study and one for the replication, to predict students' performance on the outcome variable of interest in this study.