State-Space Modeling for Binomial Time Series with Excess Zeros

• Activity Number: 310
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Business and Economic Statistics Section
• Author(s): Fan Tang* and Joseph Cavanaugh
• Companies: The University of Iowa and The University of Iowa
• Keywords: state-space models ; zero-inflation ; count time series ; particle filtering and smoothing

Abstract

Count time series are commonly encountered in many biomedical, epidemiological, and public health applications. In principle, such series may feature both zero-inflation and serial correlation. To effectively model count time series arising from a zero-inflated binomial mixture distribution, we propose a general class of parameter driven models, formulated in the state-space setting. For estimation, we develop a Monte Carlo Expectation-Maximization (MCEM) algorithm. Particle filtering and particle smoothing methods are employed to approximate the high-dimensional integrals in the E-step of the algorithm. The finite sample distributional properties of the parameter estimators are investigated through a comprehensive simulation study. Our methods are sufficiently general to accommodate various count data scenarios, and outperform comparable Poisson-type models when the data are generated from a zero-inflated binomial mixture. We illustrate the proposed methodology with a practical application.