ABSTRACT
This chapter describes to the size of a population; however, the information about that population is imperfect. However, it is generally agreed that these assumptions are unlikely to hold in human populations. Including covariates in log-linear models of population registers improves population-size estimates for two reasons. Firstly, it takes account of heterogeneity of inclusion probabilities over the levels of the covariate; and secondly, it subdivides the estimated population by the levels of the covariates, giving insight into characteristics of individuals that are not included in any of the registers. The chapter discusses the invariance of population-size estimates derived from log-linear models that include covariates, with the same covariates available in all the registers. Creating a population breakdown by passive covariates through an appropriate collapsible model is an elegant way to tackle this important practical problem. The chapter considers how the existence of invariant population sizes can help in choosing an appropriate model.
