Off-campus UMass Amherst users: To download dissertations, please use the following link to log into our proxy server with your UMass Amherst user name and password.
Non-UMass Amherst users, please click the view more button below to purchase a copy of this dissertation from Proquest.
(Some titles may also be available free of charge in our Open Access Dissertation Collection, so please check there first.)
Comparison of calibration and discrimination in non-nested binary regression models
Abstract
Whether researchers are attempting to identify a subset of covariates that best explains the variability in a binary outcome, or whether they need to select from amongst previously established models the one that most accurately represents their research setting, many important applications can be framed in terms of comparison of non-nested binary regression models. We propose an approach which revolves around explicit and independent comparison of two fundamental dimensions of binary model performance; calibration and discrimination. The methods we develop compare intuitively reasonable representations of each of these two dimensions measured on two candidate models. Also, we propose classifying possible applications for these methods into one of two scenarios; development or validation. These two scenarios each warrant use of different fundamental assumptions which effect the analytic nature of the resulting test statistics. A case study is presented in which two models are first compared in the development scenario, during which the models are estimated, and second in the validation scenario in which the parameters from the previously estimated models are applied to a dataset independent of that used to estimate the models. The methods allow computationally simple statistical inference about a comparison which previously could only be made using casual comparison of the individual model's performance.
Subject Area
Biostatistics|Statistics
Recommended Citation
Chasan-Taber, Scott Michael, "Comparison of calibration and discrimination in non-nested binary regression models" (1994). Doctoral Dissertations Available from Proquest. AAI9420608.
https://scholarworks.umass.edu/dissertations/AAI9420608