Simultaneously Propagating Uncertainty and Providing Inference on Linked Data via Bayesian Graphical Record Linkage

• Activity Number: 112
• Type: Invited
• Date/Time: Monday, August 10, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Government Statistics Section
• Author(s): Andrea Tancredi* and Rebecca Steorts and Brunero Liseo
• Companies: Sapienza University of Rome and Duke University and Sapienza University of Rome
• Keywords: Bayesian Inference ; Data matching ; Linear model ; MCMC ; Matching uncertainty ; Record linkage

Abstract

We review advances on general Bayesian methods for performing record linkage across multiple databases and making inference using the resulting matched records. In particular, we cluster records to hypothesize latent entities using a formal hierarchical Bayesian model. Our novelty, is the model comprises both the matching features and the other features included for inferential aims. Thus, the pattern of links between records which are directly linked to latent true individuals is represented as a bipartite graph, in which records are directly linked to latent true individuals, and only indirectly linked to other records. Using such an approach, we can account for the matching uncertainty in inferential procedures, (e.g., multiple linear regression, capture-recapture) based on probabilistically linked data, and at the same time, we can generate a feed-back propagation of the information between the working statistical model and the record linkage process. We illustrate our methods on simulated and real data.