Competency graphs are not influenced by demographic missing data, but any later association analyses are.
A student is considered to have achieved a competency if they score a 5 or higher on that competency. Students could have scored 1 to 8, and were scored by their instructor using a common rubric. In this way, students either achieve or do not achieve a competency, a binary outcome. Therefore, we would be interested in estimating the proportion of students achieving each competency in each semester, regardless of the course or instructor.
Students belong to specific semester and instructor-course group.
This report focuses on the “Investigación científica” domain, first considering its five competencies and then domain overall.
Resampling at the instructor–course (cluster) level within each semester make the SEs and CIs robust to within-instructor correlation and heteroskedasticity. This ensures that the CIs better reflect the uncertainty in the estimates of the percentage of students achieving the given competency/domain.
Approach: cluster bootstrap percentile CI of the student-weighted proportion per semester