Estimation of finite population in varying probability sampling
by
Raghunath Arnab
Department of Statistics,University of KwaZulu-Natal, South Africa and University of Botswana, Botswana

In large-scale multi-character surveys, we very often estimate population parameters for more than one character at a time. A sample selected with varying probabilities using a single auxiliary variable may provide efficient estimators for those characters which are well related to the auxiliary variable but not provide efficient estimators for the characters poorly related to the study variable. In this paper we have proposed a few alternative estimators for estimating the population total when the study variable is poorly related to the auxiliary variable. The proposed estimators are found to be superior to the conventional estimators under certain superpopulation model.

Date received: July 13, 2004