ABSTRACT
This chapter considers log-linear modelling of list-survey capture data under the following set-up. The first one is derived from the standard log-linear models of the cross-classified cells of a contingency table, which is based on the conditional independence type of assumptions. The second class are log-linear models of error probabilities in the marginally classified list domains. A. M. Coumans estimate homelessness in the Netherlands based on capture-recapture data, under the standard log-linear models which assume that erroneous enumeration is absent. However, in many situations, erroneous enumeration, or spurious out-of-scope records, is unavoidable. Each test can have false positive and false negative results, where the former represents erroneous enumeration of the target patient population. Exploratory analysis indicates that there is room for considering models that allow for erroneous enumeration. There is little difference between the log and logit models when either the GBA or LADIS register is treated as the enumeration survey.
