Assessing Model Fit for Ordinal Response Regression Models

• In Event: Mediation, Ordinal, and Nonlinear Modeling

Abstract

Ordinal regression models (ORMs) are being used with increasing frequency in education, health, and social/behavioral research. However, there seems to be a lag in the use of strategies for assessing the quality of fit for these kinds of models. We begin by reviewing ordinal regression models and their connection to logistic regression. We then describe approaches to establishing goodness-of-fit for logistic models and, by extension, to ordinal models. Next, we present a synthesis of the research literature on simulation studies investigating goodness-of-fit for ordinal regression models. Finally, we turn to an applied example to demonstrate application of goodness-of-fit measures when outcomes are ordinal. Our demonstration illustrates the limitations of existing measures of goodness-of-fit in ORMs.

Authors

• Ann A. O'Connell, The Ohio State University

• Sandra Reed, The Ohio State University

• Sui Huang, The Ohio State University

• DeLeon Lavron Gray, North Carolina State University