In this project we develop different methodologies for estimating the regression parameters under the assumptions that the response and predictors are not jointly observed but are brought together via an error-prone record linkage process that possibly results in mismatching.