On the enhancement of estimator efficiency of population variance through stratification, transformation, and formulation with application to COVID-19 data

The exploration of efficient and reliable data analysis tools is a constant endurance in statistical community. Stratification bring enhancement in estimates by capturing the heterogeneity in the data. This work introduces a novel data-driven machine learning algorithm aiming stratification problem, formally based on subjective approach. The development of efficient variance estimators of finite population and the exploration of using various transformation to auxiliary variable on precision enhancement of variance estimators is also under consideration in this research. We also examine the improvement in efficiency enhancement using various transformations and build superiority space for each of these transformations. These superiority regions offer significant understandings of the precise and accurate conditions favoring one transformation over another. We carefully investigate the theoretical basis of the proposed estimators, defining the superiority space for each transformation and obtained biases and mean square errors up to the first-order approximation. We conduct simulation studies and empirical analysis using COVID-19 data and artificial data to assess and validate our methods thoroughly. The results clearly show that the proposed variance estimatoElbow-Methodrs perform significantly better than the competing estimators. Further, the proposed estimator can be seamlessly adapted in other sampling designs as well as in efficient parameter estimation.