Abstract

In many large scale sample surveys, usually, it is of interest to estimate parameters relating to several characteristics. Some of the study variables may be poorly correlated with the selection probabilities used, when the selection of units is done by a probability proportional to size (pps) sampling method. When such a situation of poor correlation exists, Amahia, Chaubey and Rao (ACR, 1989) proposed alternative estimators for the population total Y of the study variable y following Rao (1966) and Bansal and Singh (1985). It is established therein that the bias of the ACR estimator is smaller than the biases of the alternative estimators suggested. ACR also showed that their estimator has smaller variance compared to the conventional estimator under a usual super population model.

In this paper, our parameter of interest is mainly B, the bias of ACR estimator and our objective is to first discuss alternative estimators for estimating this bias when sampling is done with probability of selection of units proportional to the size measure. Next we shall deal with similar other situations where alternatives are to be sought, when the design used is the often adopted conventional Simple Random Sampling (SRS) design.

Author: T. J. Rao

Received on: January, 2008

Accepted on: September, 2009