Mercoledì 23, ore 12:30 presso l'Aula Fanfani (Sapienza Università di Roma, Via del Castro Laurenziano 9, Roma - Facoltà di Economia, 5°piano)
Il prof. A. Farcomeni
terrà un seminario dal titolo: "Recapture models with arbitrary list dependence, observed and unobserved heterogeneity; with application to prevalence of Italian heroin users”
Abstract:
We propose a flexible class of models for population size estimation in capture-recapture studies. Equality constraints for the conditional capture probabilities can be used to impose any in-homogeneity/dependence assumption. Observed and unobserved heterogeneity are dealt with by means of a logistic parameterization. In order to explore the extremely rich model class we introduce a penalized version of the likelihood. The conditional likelihood and penalized conditional likelihood are maximized by means of efficient EM algorithms. We are motivated by a real data application about the number of heroin users in Italy in 2008, based on four lists. We show that overlooking the unusual dependence/in-homogeneity structure, important covariates, or unobserved heterogeneity would lead to substantially biased estimates