Methods for adjusting for lack of independence in an application of the Fellegi-Sunter model of record linkage - ARCHIVED

Let A x B be the product space of two sets A and B which is divided into matches (pairs representing the same entity) and nonmatches (pairs representing different entities). Linkage rules are those that divide A x B into links (designated matches), possible links (pairs for which we delay a decision), and nonlinks (designated nonmatches). Under fixed bounds on the error rates, Fellegi and Sunter (1969) provided a linkage rule that is optimal in the sense that it minimizes the set of possible links. The optimality is dependent on knowledge of certain probabilities that are used in a crucial likelihood ratio. In applying the record linkage model, an independence assumption is often made that allows estimation of the probabilities. If the assumption is not met, then a record linkage procedure using estimates computed under the assumption may not be optimal. This paper contains an examination of methods for adjusting linkage rules when the independence assumption is not valid. The presentation takes the form of an empirical analysis of lists of businesses for which the truth of matches is known. The number of possible links obtained using standard and adjusted computational procedures may be dependent on different samples. Bootstrap methods (Efron 1987) are used to examine the variation due to different samples.