Generalized Poisson Hurdle Model for Count Data and Its Application in Ear Disease.

Area of Science:

• Biostatistics
• Epidemiology
• Statistical Modeling

Background:

• Count data often exhibit complex characteristics, including excess or insufficient zeroes (zero-inflation/deflation) and dispersion issues where variance does not equal the mean.
• Existing statistical models, such as zero-inflated and generalized Poisson models, often struggle to simultaneously address both zero-related issues and dispersion problems.
• Ear diseases represent a significant public health concern and frequently present count data with these challenging statistical properties.

Purpose of the Study:

• To introduce a novel statistical model, the generalized Poisson Hurdle model, capable of handling count data with both zero-inflation/deflation and over/under-dispersion.
• To establish the theoretical properties of the proposed model, including the asymptotic normality and efficiency of its parameter estimators.
• To validate the practical utility and performance of the generalized Poisson Hurdle model using real-world ear disease data.

Main Methods:

• Development of a generalized Poisson Hurdle model designed to accommodate simultaneous zero-inflation/deflation and dispersion.
• Application of the generalized method of moments (GMM) for parameter estimation within the proposed model.
• Theoretical analysis to establish the asymptotic normality and efficiency of the GMM estimators.

Main Results:

• The proposed generalized Poisson Hurdle model successfully handles count data with both excess/few zeroes and unequal sample variance to mean ratios.
• Parameter estimation using the generalized method of moments yielded asymptotically normal and efficient estimators.
• Application to ear disease data demonstrated that the generalized Poisson Hurdle model outperformed both the generalized Poisson model and the standard Hurdle model.

Conclusions:

• The generalized Poisson Hurdle model offers a robust and effective solution for analyzing complex count data prevalent in healthcare, such as ear disease incidence.
• The model's ability to simultaneously address zero-inflation/deflation and dispersion provides a significant advancement over existing methods.
• The established theoretical properties and superior empirical performance highlight the model's potential for broader application in biostatistics and epidemiology.