© 2021 Proteus. All rights reserved :: Website by Myth

Why I don't use AUC to assess model fit.

Area under the curve (AUC) of the receiving operator characteristic (ROC) is commonly used as a method to assess how appropriate a model is for binary data, in particular. For example, to assess how well a logistic regression analysis is at correctly classifying the binary observations. By extension, AUC often quoted as a measure of model fit.

In this video Darryl points out why he thinks AUC is of limited usefulness for assessing the absolute fit of a model, and so why he doesn't use it.