A Zero-Inflated Poisson Mixed Latent Model for Multivariate Zero-Inflated Count Data

• Activity Number: 136
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Resmi Gupta* and Yue Zhang and Maurizio Macaluso and Nanhua Zhang
• Companies: Cincinnati Children's Hospital Medical Center and University of Cincinnati and Cincinnati Children's Hospital Medical Center and Cincinnati Children's Hospital Medical Center
• Keywords: Multivariate Zero-inflated modeling ; Zero-inflated Poisson mixed latent model ; Zero-inflated count data

Abstract

It is common to observe multivariate zero-inflated count data over time in behavioral research. To analyze the psychometric properties of multivariate zero-inflated (ZI) count items measured over time, and to predict person trait scores, a ZI-Poisson mixed regression with multidimensional latent trait is proposed. Individual heterogeneity arising from successive measurement of items has been taken into account in the modeling framework. We illustrate the proposed model using a dataset from a study to assess the problems with condom use (slippage, spilling etc.) among women at high risk for sexually transmitted diseases. Probability of having problems with slippage is higher compared to other types of problems reported. Moreover, the expected frequency for problems due to slippage is higher compared to spilling or other problems. Model also demonstrates a decreasing trait in problems reported with condom use. The parameters are estimated using Bayesian inference using Markov chain Monte Carlo procedure.