Testing for Block Effects in Regression Models Based on Survey Data

Abstract

Abstract This article considers the problem of testing for intrablock or intracluster correlation in regression disturbances that may occur when cluster or two-stage sampling data is used in regression analysis. It points out that the one-sided Lagrange multiplier test is locally best invariant. An empirical power comparison suggests that if the block structure is known this test should be used. Otherwise the Durbin—Watson test provides a useful test, especially in large samples.