ABSTRACT
In this chapter we use the techniques of Exploratory Data Mining (EDM) to identify “academic risk” among college students. Many of the specific analyses described below used data from the original Academic Performance Study (APS) of Division I student-athletes collected by the National Collegiate Athletic Association (NCAA). We present two examples of how we can define academic risk, but each example offers slightly different analytic challenges. For example, there is a very limited set of variables to craft a national initial eligibility standard. Nevertheless, there are many options to consider for combining scores and weighting the variables with utility weights that are defined to set rational cut-scores. Standard modeling approaches handle some aspects of the problem well (variable weighting), but others are challenging (evaluating the accuracy of selection when rule type, cut-scores, and utilities vary simultaneously). Non-standard exploratory approaches were then designed to address these challenges in a sufficiently nuanced manner given the high stakes to both the prospective student-athletes and the colleges. Although EDM techniques such as Decision Tree Analysis (DTA) can be adapted for these purposes, there is room for additional method development in dealing with this sort of problem.
