![IconGems-Print](images/IconGems-Print.png)
506 – Categorical Data
An Overview of the Assessment of Logistic Regression Models
Timothy J. Robinson
University of Wyoming
Justin Shang
University of Wyoming
Shaun S. Wulff
University of Wyoming
The logistic regression model is frequently used in many practical applications to fit a binary response. Model specification depends upon a number of issues including response selection, link specification, and the choice of predictors. Model evaluation includes model selection, predictive ability, and goodness-of-fit. As a result, the art of logistic regression modeling involves many choices and multiple criteria for the data modeler to consider. Particular emphasis will be given to a thorough review of the model selection procedures and the goodness-of-fit testing. In logistic regression, goodness-of-fit assessments sometimes can be challenging, depending on the covariates in the model and the number of covariate patterns. Goodness-of-fit tests can involve chi-square based tests, raw residuals, and transformed residuals. We detail these approaches for assessing the quality of logistic regression models.