A NEW CLASS OF REGRESSION ESTIMATOR FOR THE POPULATION MEAN USING RANKED SET SAMPLING

Abstract

Ranked Set Sampling (RSS) was initially suggested to increase the e ciency of the population mean. It has been shown that this method is highly bene cial to the estimation based on simple random sampling (SRS). The problem of estimating the population mean is an integral aspect of scienti c survey. More e cient and less biased estimators in the form of regression estimator with minimum mean square errors were obtained for population means. The estimator was examined for both cum-dual product and cum-dual ratio under ranked set sampling without replacement. Expressions for the mean, mean square error (MSE), bias and variance of the proposed estimators were derived to rst order of approximation. The MSE of the newly proposed estimator for the product case was better for the correlation coe cient in the range [-0.1 to 0.2] while for the ratio case, the MSE was more e cient for the correlation coe cient in the range [-0.1 to 0.3] when compared to other existing estimators using monte carlo simulation. These results indicate that the proposed new class of regression estimator for the population mean using ranked set sampling is more e cient when compared to estimators based on simple random sampling (SRS) and some existing estimators based on RSS.