Thomas Lumley • 1/6/2023

Linkage and multiple imputation

This technical article discusses extending multiple imputation techniques to the problem of probabilistic record linkage. It explains the classical Fellegi & Sunter framework, critiques single imputation of links, and proposes a Bayesian approach to sample multiple plausible sets of record links. This allows analysts to propagate linkage uncertainty into final models using standard multiple imputation combining rules, separating linkage expertise from analysis.

0 comments

#Data Science #Statistical Modeling #Bayesian Inference