New Class of Estimators for Finite Population Mean Under Stratified Double Phase Sampling with Simulation and Real-Life Application

Sampling survey data can sometimes contain outlier observations. When the mean estimator becomes skewed due to the presence of extreme values in the sample, results can be biased. The tendency to remove outliers from sample data is common. However, performing such removal can reduce the accuracy of conventional estimating techniques, particularly with regard to the mean square error (MSE). In order to increase population mean estimation accuracy while taking extreme values into consideration, this study presents an enhanced class of estimators. The method uses extreme values from an auxiliary variable as a source of information rather than eliminating these outliers. Using a first-order approximation, the properties of the suggested class of estimators are investigated within the context of a stratified two-phase sampling framework. A simulation research is conducted to examine the practical performance of these estimators in order to validate the theoretical conclusions. To further demonstrate the superiority of the suggested class of estimators for dealing with extreme values, an analysis of three different datasets demonstrates that they consistently provide higher percent relative efficiency (PRE) when compared to existing estimators.