Online Program Home
My Program

Abstract Details

Activity Number: 375
Type: Contributed
Date/Time: Tuesday, August 2, 2016 : 10:30 AM to 12:20 PM
Sponsor: Section on Statistics in Epidemiology
Abstract #319527
Title: Maximum Likelihood or Generalized Estimating Equations: A Comparison in the Context of Proportional Odds Model for Ordinal Response
Author(s): Xinkai Zhou*
Companies: Statistics Core@UCLA
Keywords: Ordinal data ; Proportional Odds model ; Maximum Likelihood (ML) ; Generalized Estimating Equations (GEE) ; Longitudinal data ; Simulation

For analyzing ordinal response data, a popular approach is to use proportional odds model and estimate model parameters by Maximum Likelihood (ML). As an alternative, Clayton (1992) proposed to re-code ordinal response as correlated binary responses and estimate parameters by Generalized Estimating Equations (GEE). This work conducts a comprehensive comparison between the two approaches. The main contribution is the simulation result comparing the bias and efficiency of parameter estimates under scenarios commonly encountered in biomedical applications: small sample size or imbalance in response distribution. In addition, we show that the working correlation matrix from Clayton's approach is consistent and asymptotically efficient under latent variable structure. We also provide R functions that reshape an ordinal response dataset to the format required by Clayton's approach. This work provides valuable information and computer programs for those who work with ordinal response data.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2016 program

Copyright © American Statistical Association