A Spline-Based Lack-of-Fit Test for Parametric Zero-Inflated Poisson Regression Models

• Activity Number: 504
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Chin-Shang Li*+
• Companies: University of California at Davis
• Keywords: B-splines ; expectation-maximization algorithm ; likelihood ratio test ; semiparametric ZIP regression ;

Abstract

Zero-inflated Poisson (ZIP) regression model have been popularly used for analysis of count data containing many zeros from biomedical studies, criminology, environmental economics and others. In a parametric ZIP regression model, a linear predictor is usually used for the effect of a covariate. To assess the validity of the linear relationship, cubic B-splines are used to approximate the covariate effect. The proposed semiparametric ZIP regression model can greatly enhance modeling flexibility. The semiparametric model parameters are estimated by maximizing the likelihood function through an expectation-maximization algorithm. A likelihood ratio test is then used to assess the adequacy of the linear relationship. A simulation study is conducted to study its power performance. A real example is provided to illustrate the practicality of the methodology.