Estimating First-Year Student Attrition Rates: An Application of Multilevel Modeling Using Categorical Variables

There have recently been significant theoretical developments in multilevel statistical modeling, and improved software is readily available. This study demonstrates the application of multilevel modeling to one of the most common issues that confront institutional researchers: that of student attrition, where the response variable is typically binary rather than continuous. Comparisons are made with a traditional logistic regression approach. The data pertain to one large university. The techniques illustrated may be extended to the analysis of data sets encompassing many institutions, making meaningful interinstitutional comparisons of performance feasible even when there is hierarchical clustering present in the data.