Minimum MSE regression estimator with estimated population quantities of auxiliary variables

Abstract

Construction of a regression estimator in which the population means of auxiliary variables are estimated with a larger sample is considered. Using the variances of the estimated population means, and the correlation between auxiliary variables and the variable of interest, a design consistent regression estimator that has minimum model mean squared error under a working model is derived. A limited simulation study shows that the minimum model mean squared error regression estimator performs well compared to the generalized least squares regression estimator, even when the working model is inappropriate. (C) 2008 Elsevier B.V. All rights reserved.