Online Program Home
My Program

Abstract Details

Activity Number: 549
Type: Contributed
Date/Time: Wednesday, August 3, 2016 : 10:30 AM to 12:20 PM
Sponsor: Survey Research Methods Section
Abstract #320619 View Presentation
Title: Modeling Similar Nonmatches in Record Linkage with Mixture Models
Author(s): Michael Larsen*
Companies: The George Washington University
Keywords: File matching ; Fellegi-Sunter ; False Match ; False nonmatch ; Data fusion ; Latent Class Model
Abstract:

In the task of record linkage, one compares information on records in two files on a single population and decides which of the pairs of records, one from each file, pertain to single individuals. The result of record linkage is a file with one record per individual that contains the sum of all information on the individuals from the two files. One also learns the number of records on the two files that appear on both and on one but not the other. Probabilistic record linkage uses statistical models to estimate the probability that a pair of records arises from a single person. Several fields of information, including names, dates, locations, and relations to other records, on the two files are compared. One set of models used to estimate probabilities based on these data are latent class and mixture models. After linkage one often desires to estimates relationships among variables that were included on the two files. Linkage is necessary when analysis variables are separated across the two files. Enhanced mixture models are proposed and illustrated for the case in which some nonmatches are similar to one another due to constraints in blocking criterion used to screen pairs.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2016 program

 
 
Copyright © American Statistical Association