A Latent Mixture Approach to Modeling Zero-Inflated Bivariate Ordinal Data

• Activity Number: 251
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Rajendra Kadel*+ and Gatachew Dagne
• Companies: Univ of South Florida and University of South Florida
• Keywords: bivariate ; Zero-inflated ; Ordinal Response ; probit model ; MCMC ; Bayesian

Abstract

Multivariate ordinal response data such as severity of pain, degree of disability, satisfaction with healthcare provider are prevalent in many areas of research including public health-related, biomedical and social science research. Ignoring the multivariate features of the response variables gives a wrong inference. In addition, such multivariate ordinal outcomes very often exhibit excess zeros at the lower end of the scales. Excess number of zeros in ordinal data coupled with multivariate structure makes it difficult to analyze and interpret. Traditional logit or probit models have limited capacity to explain such zero-inflation and the estimates of the parameters are likely to be biased or numerical convergence may fail. Existing methods for the zero-inflated data are limited to univariate-logit or univariate-probit model and extension to bivariate (or multivariate) models has been very limited. We propose a mixture approach to modeling excess zeros in bivariate ordinal data using. A latent variable approach will be used to develop a Mixture Bivariate Zero-Inflated Ordered Probit (M-BZIOP) model. We will employ Bayesian MCMC technique for the estimation of the parameters.