Ordinal Regression Analysis: Unconstrained Continuation Ratio Models and Partial Continuation Ratio Models for Ordinal Response Variables

• In Event: Roundtable Session 31
  In Roundtable Session: Alternatives to Parametric Analysis in the Contexts of Regression and Analysis of Variance

Abstract

In addition to the proportional odds (PO) model, the continuation ratio (CR) model is another well-known technique for ordinal response variables when our interest focuses on a particular category rather than at or below that category in the PO model. The CR model also assumes the PO assumption or parallel lines assumption. However, this assumption is often violated. The purpose of this paper is to illustrate the use of the unconstrained and partially constrained CR models to estimate educational data with Stata, and compare the results of fitting these two models and the CR model with the PO assumption. For demonstration purposes, the High School Longitudinal Study of 2009 (HSLS: 2009) data were used to conduct the ordinal regression analyses.

Authors

• Xing Liu, Eastern Connecticut State University

• Hari P. Koirala, Eastern Connecticut State University