Online Program

Modeling Assessment Data with a Hierarchical Approach (303042)

Keywords: assessment data, hierarchical models, clustering, data analysis, shiny

Assessment data collected on student conceptual performance and attitudes in a randomization-based statistics curriculum during fall 2013 and spring 2014 were analyzed using a hierarchical modeling technique. Both a concepts and attitudes test was administered to students at the start and end of the course. Therefore, the goal of this research was to detect the characteristics of students and instructors that were significantly associated with student conceptual gains after undergoing this type of curriculum. Due to the large number of variables in the data set, we conducted k-means and hierarchical clustering to distinguish instructors separately based on the types of students entering the course and the structure of the course. We then ran multiple hierarchical models before producing a final model. To communicate our findings and the flaws of neglecting hierarchy, we developed a Shiny app. One feature of the Shiny app is that it illustrates the idea of shrinkage in the parameter estimates with a hierarchical model. We also allow users to analyze their own data and visualize how a pooled, unpooled, and hierarchical modeling method differs.