Augmented Marginal Log-linear Models for Capture-recapture Studies

• Activity Number: 50
• Type: Contributed
• Date/Time: Sunday, August 8, 2004 : 4:00 PM to 5:50 PM
• Sponsor: Biometrics Section
• Author(s): Elizabeth L. Turner*+ and Alain C. Vandal
• Companies: McGill University and McGill University/SMBD Jewish General Hospital
• Keywords: capture-recapture ; source dependence ; conditional independence ; augmented marginal log-linear models ; epidemiology

Abstract

In capture-recapture modeling it is necessary to account for possible dependence between sources. We propose a novel technique, termed augmented marginal log-linear modelling (AMLLM), to account for both heterogeneity-induced and pure source dependence in capture-recapture studies when individual covariate data are collected. Central to this technique, under the assumption that sources are conditionally independent given the covariates, is a measure we call the coefficient of source dependence (CSD), defined for every set of sources. The CSDs are formed by two components: the distribution of the covariates in each source and the population distribution of the covariates. The first is estimated empirically and the second estimated along with the unknown population size by inclusion of the CSDs in the AMLLM. This technique permits, within a marginal log-linear model, the inclusion of covariate data and the estimation of the population covariate distribution and population size. It also avoids the problems of model selection at source level and random zeros. We illustrate the use of CSDs in the design of epidemiological capture-recapture studies.