Bayesian Partitioning Approaches to Record Linkage Uncertainty by Samuel Wolf ’21

Bayesian Partitioning Approaches to Record Linkage Uncertainty by Samuel Wolf ’21, Statistics Colloquium, Monday, April 12, 1:30 – 2:15 pm, meet at https://meet.google.com/abm-rhgs-fmb.

Abstract:  In this colloquium, I discuss a Bayesian approach to record linkage, or the process by which records from disparate data sources are linked together.  Record linkage often suffers from uncertainty over whether or not two records are “coreferent,” or refer to the same individual.  A novel approach by Mauricio Sadinle (2014) uses Bayesian principles to assess the probability of coreference partitions by detecting potential duplicates.  This methodology is then applied to incomplete and error-ridden homicide records from the El Salvadoran civil war.