Skip to main content

Advertisement

Table 4 Summary of statistical tests used

From: Observing populations and testing predictions about genetic drift in a computer simulation improves college students’ conceptual understanding

Question motivating the model Model Predictor Estimate ± SE p value Result
Do the students in the two treatments perform differently before instruction? pretest performance ~ treatment + course + OLRE Intercept 0.13 ± 0.17 0.427 No. The students in the two treatments did not differ in pretest performance
Treatment 0.32 ± 0.19 0.094
Does treatment predict posttest performance (controlling for pretest raw score)? posttest performance ~ pretest raw score + treatment + course + OLRE Intercept −1.19 ± 0.23 <0.001 Yes. The students in the module courses outperformed students in the control courses
Pretest 0.11 ± 0.01 <0.001
Treatment 0.83 ± 0.24 <0.001
Does treatment predict performance on items about key concepts (controlling for pretest raw score)? posttestKC performance ~ pretestKC raw score + treatment + course + OLRE Intercept −0.69 ± 0.33 0.036 Yes. Students in the module courses outperformed students in the control courses
PretestKC 0.09 ± 0.03 <0.001
Treatment 1.92 ± 0.35 <0.001
Does treatment predict performance on items about misconceptions (controlling for pretest raw score)? posttestM performance ~ pretestM raw score + treatment + course + OLRE Intercept −1.08 ± 0.26 <0.001 Yes. Students in the module courses outperformed students in the control courses
PretestM 0.17 ± 0.01 <0.001
Treatment 0.56 ± 0.24 0.018
  1. In all models, Course is a random factor that accounts for differences across classes that were not controlled during the study. Observation-level random effect (OLRE) is a random factor that accounts for overdispersion in the model. Subscript KC indicates the subset of items in the GeDI about key concepts; subscript M indicates the subset of items on the GeDI about misconceptions. Significant results after Holm-Bonferroni corrections are in italics