Assessing cumulative logit models via a score test in random effect models

Abstract

The purpose of this article is to develop a goodness-of-fit test based on score test statistics for cumulative logit models with extra variation of random effects. Two main theorems for the proposed score test statistics are derived. In simulation studies, the powers of the proposed tests are discussed and the power curve against a variety of dispersion parameters and bandwidths is depicted. The proposed method is illustrated by an ordinal data set from Mosteller and Tukey [23].

Keywords:

• cumulative logit models
• local linear smoother
• ordinal score
• power
• score test

Acknowledgements

I am grateful to the Editor and reviewers for their valuable comments and constructive suggestions on the original draft, which substantially led to improvements in the paper.