Before a regression model is used to address questions about the relationship between a response variable and predictors, the fit of the model to the data should be assessed. For example, consider a logistic regression model for explaining the dependence of a binary outcome variable on a set of predictor variables or for predicting the outcome variable based on the predictors. Checking the fit of the model before it is used in a practical setting is of critical importance. If a model is found to be deficient, the nature of the deficiency may indicate a need for some aspect of the model to be reformulated or that poorly fitting observations need to be considered separately. I propose graphical methodology based on a Bayesian framework to help address issues such as this. Plots can be constructed quickly and easily for any model of interest, and goodness of fit assessed. These plots are more intuitive and easy-to-use than traditional graphical diagnostic methods for regression such as residual plots.
Back to Research page.
Send me e-mail at ipardoe at lcbmail.uoregon.edu
Last updated: October 8, 2001
The views and opinions expressed in this page are strictly those of the page author. The contents of this page have not been reviewed or approved by the University of Oregon.
© 2001, Iain Pardoe, Lundquist College of Business, University of Oregon