On optimal classes of estimators in the presence of some non-sampling errors

During a survey study, an investigator may be incapable of assembling the complete response (i.e., there is non-response) and/or the assembled response is not 100 % true (i.e., measurement errors exist). In this situation, estimation of the population mean under stratified random sampling is not an easy task. Mostly these non-sampling errors, i.e. non-response and measurement error, significantly affect the estimators than sampling errors. To deal with this task, a progressive generalized estimator has been proposed, that can generate a number of estimators based on the availability of conventional and/or non-conventional auxiliary information. Ratio-type, ratio-type exponential, ratio-ratio-type exponential, ratio-product-type exponential, product-type, product-type exponential, product-product-type exponential and product-ratio- type exponential estimators are generated through the proposed generalized estimator. Mathematical properties such as bias, mean squared error and minimum mean squared error of the proposed estimator are derived up to first degree of approximation. The empirical performance of all the estimators in terms of percent relative efficiency is evaluated with the help of a simulation study. It turned out that the proposed estimators outperform when compared with Hansen and Hurwitz (1946) estimator and other competing estimators in this study i.e. Singh and Kumar’s (2008), Kumar et al. (2015), Azeem and Hanif (2017) and Zahid and Shabbir (2018). It is suggested that the proposed estimators will be applied in case of non-response and measurement errors under stratified random sampling.