Alternative estimators for multi-character surveys in complex survey designs
by
Raghunath Arnab
Univesity of Durban-Westville, South Africa

The problems of estimating finite populatin mean in multi-character surveys are considered in a unified set up. It is shown that under certain superpopulation model, the sample mean fares better than any other estmator belong to the class of linear homogeneous unbiased estimator for the population mean. Alternative estimators for various sampling designs are proposed when measure of size is not well related to the study variable. Sufficient conditions for superiority of the proposed estimators over the conventional Horvitz Thompson estimator under superpopulation model have been derived. Numerical illustrations are also presented to study the performances of the proposed estimators.

Date received: October 29, 2002