Empirical Likelihood Approaches under Complex Sampling Designs

Abstract

Abstract There are two different empirical likelihood approaches for complex sampling designs: “ pseudoempirical likelihood ” and “ unequal probability empirical likelihood ”. Both approaches are described and reviewed critically. The key difference is the fact that the self‐normalization property of the pseudoempirical likelihood approach is limited to unidimensional parameters. This property holds for multidimensional parameters, with the unequal probability empirical likelihood approach. This manuscript is a brief description of the key empirical likelihood approaches for complex sampling. This is not an exhaustive account of all the applications of empirical likelihood in survey sampling.