|
Activity Number:
|
22
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, July 29, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #309966 |
|
Title:
|
Estimating Kappa Coefficient and Tetrachoric Correlation for Clustered Binary Data
|
|
Author(s):
|
Xiao Zhang*+ and Gary Cutter
|
|
Companies:
|
The University of Alabama at Birmingham and The University of Alabama at Birmingham
|
|
Address:
|
1665 University Blvd. 414A, Birmingham, AL, 35294,
|
|
Keywords:
|
Kappa Coefficient ; Tetrachoric Correlation ; Multivariate Probit Model ; Markov chain Monte Carlo ; Simulation Study ; Sensitivity
|
|
Abstract:
|
We estimate the kappa coefficient and the tetrachoric correlation for clustered binary outcomes from a Bayesian perspective. We construct a Bayesian model and develop a Markov chain Monte Carlo (MCMC) algorithm to sample the tetrachoric correlation. Based on the definition of the kappa coefficient which is the function of the mean and the tetrachoric correlation of the binary outcomes, our method produces the posterior inference of the kappa coefficient. We investigate the sensitivity of the posterior estimation of the kappa coefficient and the tetrachoric correlation. We illustrate our method through the psychiatric data used by Lipsitz, Laird, and Brennan (1994) and through the data from an epitope-specific immunotherapy in rheumatoid arthritis study.
|
