A Generalized Fellegi-Sunter Framework for Multiple Record Linkage With Application to Homicide Record Systems

We present a probabilistic method for linking multiple datafiles. This task\nis not trivial in the absence of unique identifiers for the individuals\nrecorded. This is a common scenario when linking census data to coverage\nmeasurement surveys for census coverage evaluation, and in general when\nmultiple record-systems need to be integrated for posterior analysis. Our\nmethod generalizes the Fellegi-Sunter theory for linking records from two\ndatafiles and its modern implementations. The multiple record linkage goal is\nto classify the record K-tuples coming from K datafiles according to the\ndifferent matching patterns. Our method incorporates the transitivity of\nagreement in the computation of the data used to model matching probabilities.\nWe use a mixture model to fit matching probabilities via maximum likelihood\nusing the EM algorithm. We present a method to decide the record K-tuples\nmembership to the subsets of matching patterns and we prove its optimality. We\napply our method to the integration of three Colombian homicide record systems\nand we perform a simulation study in order to explore the performance of the\nmethod under measurement error and different scenarios. The proposed method\nworks well and opens some directions for future research.\n