Estimation of Location Difference for Fragmentary Samples

摘要

A class of simple and robust estimators of the difference between location parameters of correlated variables is proposed when some observations on either of the variables are missing. We show that these estimators are consistent, asymptotically normally distributed, and insensitive to outlying observations. Asymptotic relative efficiency comparisons with other known estimators are made to show the advantage of the proposed estimators. (Author)