Fig 1.
Two plots showing the effect of the matched pairs randomization design (1A) as compared with complete randomization (1B).
Each point represents a student’s covariate information, and each connecting edge indicates that those students have been assigned to opposite treatment groups. The edges in the matched pairs design are much shorter than under complete randomization, confirming that matching produces more similar randomizations.
Table 1.
The effect sizes of peer assessment in the short term and long term.
(Standard errors are shown in parentheses.)
Fig 2.
Distribution of scores for the control (blue) and treatment (red) groups on quiz 5 in the winter quarter.
The dashed vertical lines designate the means. (The difference in means on this quiz was 5.9.) Similar plots for all of the quizzes and final exam may be found in (S1 File).
Table 2.
Achievement gaps in our population of students, reported as an effect size.
We show the gap before and after the course. (Standard errors are shown in parentheses.) The “before” numbers were calculated using scores on a pre-quiz administered prior to the randomization. The “after” numbers were calculated using scores on the final exam.