Parameterization of Continuous Covariates in the Poisson Capture-Recapture Log Linear Model for Closed Populations

Abstract

The capture-recapture method is widely used by epidemiologists to estimate the size of hidden populations using incomplete and overlapping lists of subjects. Closed populations, heterogeneity of inclusion probabilities and dependence between lists are taken into consideration in this work. The main objective is to propose a new parameterization for the Poisson log linear odel (LLM) to treat continuous covariates in their original measurement scale. The analytic estimate of the confidence bounds of the hidden population is also provided. Proposed model was applied to simulated and real capture-recapture data and compared with the multinomial conditional logit model (MCLM). The proposed model is very similar to the MCLM in dealing with continuous covariates and the analytic confidence interval performs better than the bootstrap estimate in case of small sample size.

Keywords

• capture-recapture
• closed population
• continuous covariates
• poisson log-linear model
• multinomial conditional logit model