A multivariate correlated poisson generalized inverse gaussian regression model for dependent count data: Estimation...

Area of Science:

• Statistics
• Biostatistics
• Epidemiology

Background:

• Multivariate count data analysis faces challenges with overdispersion and response variable correlation.
• Existing models may not adequately address these complexities in public health contexts.

Purpose of the Study:

• To introduce a novel regression model, Multivariate Correlated Poisson Generalized Inverse Gaussian Regression (MCPGIGR), for analyzing multivariate count data.
• To provide robust estimation and hypothesis testing methods for the proposed model.
• To demonstrate the practical utility and improved model fit of MCPGIGR in a public health application.

Main Methods:

• Development of the MCPGIGR model incorporating random effects via common shock variables and a flexible log-link function.
• Implementation of Maximum Likelihood Estimation (MLE) and Maximum Likelihood Ratio Tests (MLRT) for parameter estimation and significance testing.
• Conducting simulation studies to validate estimator consistency and performance.

Main Results:

• The MCPGIGR model demonstrated a substantial improvement in model fit compared to Multivariate Poisson Regression (MPR) in an application to maternal and neonatal mortality data.
• Simulation studies confirmed the consistency and performance of the proposed estimators.
• The model effectively handles overdispersion and correlation in multivariate count data.

Conclusions:

• The MCPGIGR model offers a flexible and practical solution for analyzing correlated, overdispersed multivariate count data.
• The framework provides robust parameter estimation and hypothesis testing, enhancing interpretability in public health research.
• MCPGIGR represents a significant advancement for statistical modeling in epidemiology and related fields.