Séminaire MAI

Présentation d'Eric Agnondi SIYA GNON intitulée « Marginal effects for zero-inflated semi-continuous data ».

Zero-inflated semi-continuous data (ZISCD) are common in various fields such as medicine, economics, and social sciences, where the outcome distribution consists of a point mass at zero and a positive continuous distribution. This work uses marginalized regression models (MZIG and MZILN) to analyze semi-continuous data with zero inflation. These models allow for a direct interpretation of covariate effects on the marginal mean, accounting for distinct mechanisms generating structural zeros and continuous positive outcomes.

The approach is validated through simulations and the analysis of real data, demonstrating a significant improvement in the interpretation and accuracy of estimates. This paper introduces two novel marginalized regression models, MZIG and MZILN, that address key limitations in existing methods by enabling a clear interpretation of covariate effects on the marginal mean, while offering robust statistical properties validated through simulations and real-world applications. This work fills a significant gap in the literature by offering a comprehensive approach to the estimation and interpretation of the MZIG and MZILN models.

Ce travail de thèse est réalisé sous la direction d'Essoham ALI (UCO Angers) et d'Aliou DIOP (Université Gaston Berger de Saint-Louis, Sénégal).