On the Performance of Some Regression-Type Estimators Obtained Through Controlled Splitting of Sample

Abstract

In this paper, we have developed four new regression-type estimators on the basis of controlled splitting of the selected sample under simple random sampling without replacement. For the splitting technique discussed here, the size of the original sample need not be an even number and the sizes of the two parts of the sample need not be equal. The performances of the suggested estimators have been compared empirically with the help of a simulation study in respect of the relative bias, relative efficiency, coverage rate based on 95% confidence interval, standard deviation of the student  statistic and approach to symmetry.