Séminaire MAI

Présentation d'Essoham ALI, enseignant-chercheur à l'UCO, intitulée « Regularized Estimation for Right-Censored Zero-Inflated Poisson Regression: Methods and Applications to Health Data ».

In biomedical and epidemiological studies, count data often exhibit excess zeros, right censoring, and multicollinearity among covariates, posing significant challenges to standard regression approaches and potentially compromising inference. In this work, we propose a regularized estimation framework for the right-censored zero-inflated Poisson (RCZIP) regression model. We introduce and compare three penalized estimators: Ridge, Liu, and a novel Modified Ridge-Type (MRT) estimator, designed to address multicollinearity while accounting for both zero inflation and censoring. The performance of these estimators is evaluated through extensive Monte Carlo simulations under varying levels of censoring and collinearity, using metrics such as mean squared error (MSE), mean absolute error (MAE), and mean squared deviation error (MSDE). The results show that the proposed MRT estimator consistently outperforms both the maximum likelihood estimator and existing penalized methods, particularly in settings with high censoring and strong multicollinearity. To demonstrate practical relevance, the proposed methodology is applied to two real-world health datasets: healthcare utilization data from the National Health and Nutrition Examination Survey (NHANES) and social contact data from Mayotte. These applications highlight the effectiveness of regularized RCZIP models in improving estimation accuracy and interpretability in complex biomedical data contexts.