ABSTRACT

CONTENTS 21.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353 21.2 Model Selection Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356 21.3 Ordinal Logistic Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359 21.4 Example Problem: Diabetes Severity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367 21.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369

Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 370

21.1 Introduction Ordinal logistic regression is a relatively new method now widely used for analyzing ordinal data. There has been very little work on model selection in ordinal logistic regression. In this paper we are interested in applying an informational complexity criterion for best subset selection in ordinal logistic regression models and for goodness-of-fit in proportional odds and nested cumulative link models.