A Marginalized Zero-Inflated Negative Binomial Regression Model with Overall Exposure Effects

• Activity Number: 452
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): John Preisser*+ and Kalyan Das and D. Leann Long and John Stamm
• Companies: University of North Carolina and University of Calcutta and West Virginia University and University of North Carolina
• Keywords: dental caries ; excess zeros ; marginal effects ; overdispersion ; zero-inflation

Abstract

The zero-inflated negative binomial regression model (ZINB) is frequently employed in medical and dental research to examine relationships between exposures of interest and overdispersed count outcomes exhibiting many zeros. The regression coefficients of ZINB have latent class interpretations for a susceptible population at risk for the disease/condition under study with counts generated from a negative binomial distribution and for a non-susceptible population that provides only zero counts. The ZINB parameters, however, are not well-suited for estimating the overall effects of explanatory variables in the mixture population. We propose a marginalized zero-inflated negative binomial regression model (MZINB) for independent responses that models the population mean count directly, providing straightforward inference for overall exposure effects based on maximum likelihood estimation. Through simulation studies, the performance of MZINB is compared to marginalized zero-inflated Poisson regression and negative binomial regression. The models are applied to count outcomes from a randomized clinical trial comparing mean caries increment for different toothpaste formulations.