A Pseudo-Likelihood Approach to Linear Regression With Partially Shuffled Data

Recently, there has been significant interest in linear regression the situation where predictors and responses are not observed matching pairs corresponding to same statistical unit as a consequence of separate data collection uncertainty integration. Mismatched can considerably impact model fit disrupt estimation parameters. In this article, we present method adjust for such mismatches under “partial shuffling” which sufficiently large fraction (predictors, response)-pairs their correct correspondence. The proposed approach is based on pseudo-likelihood each term takes form two-component mixture density. expectation-maximization schemes optimization, (i) scale favorably number samples, (ii) achieve excellent performance relative an oracle that access pairings certified by simulations case studies. particular, tolerate larger than existing approaches, enables noise level well mismatches. Inference resulting estimator (standard errors, confidence intervals) be established theory composite likelihood estimation. Along way, also propose test presence establish its consistency suitable conditions. Supplemental files article available online.