Riccardo Faini CEIS Seminars

Multiple Systems Estimation for Hard-to-count Populations: Some Recent Developments
October, 02nd 2020 (16:00-17:00)
TEAMS Webinar - Registration required

Sir Bernard Silverman (University of Nottingham)

Riccardo Faini CEIS Webinars

Registration form

Multiple systems estimates have been used in many contexts where direct quantification of a population is not possible, especially in the general area of human rights. They are one of the approaches used to estimate the numbers of victims of human trafficking and modern slavery, and a multiple systems estimate was a leading component of the UK Government’s strategy leading to the passing of its Modern Slavery Act. I will survey the use of the method in this context and consider some specific issues that arise. In particular, the sparseness of the data tables requires attention to aspects of generalised linear models beyond those addressed in standard software packages. Furthermore, not only for sparse data but more generally, because there are many different possible models to choose between, statistical inference should not be carried out conditional on the chosen model but should take account of the model choice step. Aspects of a bootstrap approach will be considered. It is necessary to pay careful attention to the computational burden even with modern computing power.