Biostatistics seminar
Assistant Professor Fode Tounkara will give a talk titled "Copula-Based Regression for Closed-Population Capture–Recapture with Time Variation and Residual Heterogeneity."
About
Fode Tounkara, PhD is an assistant professor in the Department of Biomedical Informatics in the College of Medicine at Ohio State.
Lunch will be provided during the seminar.
Abstract
Capture–recapture methods are widely used to estimate the size of populations that are only partially observed, with applications ranging from ecology to epidemiology and public health surveillance. A central challenge is heterogeneity in capture probabilities across individuals and sampling occasions. Regression-based capture–recapture models can account for observed individual characteristics and time variation, but commonly rely on conditional independence of repeated capture outcomes after adjustment for measured covariates. When residual heterogeneity remains, this assumption may lead to biased estimation of capture probabilities and population size.
In this talk, I will present a flexible copula-based regression framework for closed-population capture– recapture studies that jointly accommodates time-varying capture probabilities, observed individual heterogeneity, and residual dependence among repeated captures. Marginal capture probabilities are modeled through regression, while Archimedean copulas characterize the remaining dependence structure. Estimation is based on a conditional likelihood, with population size subsequently estimated using a Horvitz– Thompson approach. The framework encompasses several commonly used capture–recapture models as special cases and allows alternative dependence structures to be compared using likelihood-based criteria. Through simulation studies, I will examine the consequences of ignoring residual heterogeneity for population size estimation and uncertainty quantification, as well as the ability to identify the underlying copula structure. An application to a harvest mouse capture–recapture study illustrates the practical impact of accounting for residual dependence. I will conclude by discussing extensions and potential applications in epidemiology, disease surveillance, and public health research
Contact
Andy Ni