Analysing capture–recapture data when some variables of heterogeneous catchability are not collected or asked in all registrations

Abstract

Registrations in epidemiological studies su
er from incompleteness, thus a general consensus is to use capture–recapture models. Lately, there has been a thrust to incorporate covariates which relate to the capture probabilities in order to improve the estimate of population size. Current approaches utilize covariates observed in all the pertinent registrations. In epidemiological applications, one often encounters capture–recapture data where some covariates are measured by some but not by all of the registrations under scrutiny. In this article, the usefulness of these types of covariates is evaluated and furthermore a methodology for analysing such capture–recapture data is presented. The methodology is illustrated by applying it to data on neural tube defects in the Netherlands. Copyright ? 2006 John Wiley & Sons, Ltd. KEY WORDS: capture–recapture; covariates; EM algorithm; log-linear models; population size estimation