Record linkage has been used to combine records in separate files that relate to the same individual or household. In this article, modifications of the typical record linkage algorithm are made for a setting where links are "competitive," that is, a record can be used only once. The goal of the linkage is to recover a missing covariate for use in the logistic regression of tag retention on fish length at time of tagging. The imputation process necessarily adds measurement error to the covariate, biasing logistic regression coefficient estimates toward zero. The extent of the bias is estimated through simulations, and adjustments are made to account for the shrinkage.

The data analyzed came from a study of tag retention for visual implant tags in cutthroat trout, where the measurements taken at time of tagging were unavailable for fish that lost the single tag.

Key Words
Bias adjustment; Logistic regression; Mark–recapture; Measurement error; Statistical matching.

James F. Robison-Cox is Assistant Professor of Statistics, Department of Mathematical Sciences, Montana State University, Bozeman, MT 59717.