Abstract:
|
College football coaches at major division I schools are highly sought after and commonly among the highest paid public employees in every state. In addition, these college football coaches serve as the public face for the University, athletic program, and state. With such a high profile job, one would imagine that the hiring process would produce success stories. Often, however, this is not the case. Sports news is filled with stories about college football coaches that did not pan out, so much so, that the success is seemingly random. In this paper we use multilevel item response theory (MIRT) to estimate NCAA head football coaches' abilities. Three approaches to estimation are considered: dichotomous, polytomous, and weighting. Ability estimates will be compared across models and the strengths/weaknesses of each approach will be discussed. Lastly, conditional MIRT models are fitted to explain variation in ability level to understand factors that may underlie college football coaches' success.
|