Multiple Systems Estimation, or capture-recapture, is a statistical method that enables the estimation of a total population using a set of intersecting, partial lists. For example, this method has been used to estimate the total number of people killed by the Army in Guatemala during the 1980s using lists of deaths compiled by human rights projects. From this analysis, we found that the relative risk of being killed by the Army was eight times greater for indigenous relative to non-indigenous residents of the Ixil area.
In this workshop, I will explore the estimate of total homicides in Colombia (disaggregated over time and space) using a set of partial registries. Each individual list has only a small fraction of the total deaths, and even together they cover about two-thirds of the total.
I will trace the history of MSE from the original estimator through various improvements that have relaxed some of the assumptions and narrowed others. These include log-linear estimates, lower-bound estimates, and two Bayesian approaches.
Participants should have some understanding of regression-like models, including log-linear estimates of contingency table counts. The classic source for this is Bishop, Fienberg, and Holland, Discrete Multivariate Analysis. The examples will be presented in R.
Spaces are limited for this workshop. Registration will close at 17:00 on Thursday 15 September 2016.