An Empirical and Simulation-Based Evaluation of Existing Class Estimators in Two-Occasion Successive Sampling

Abstract

This study presents an empirical and simulation-based comparison of four established estimators for estimating the population mean in two-occasion successive sampling. Artificial populations have been generated under varying correlation structures (strong, moderate, and weak) and different sample sizes to evaluate estimator their performances using percent relative efficiency (PRE) and the optimum replacement policy. The results reveal that estimators’ efficiencies increase with increase in correlation strength and sample size. Real-data applications supported the simulation outcomes, confirming the superior and consistent performance of some estimators over others across multiple populations. Overall, no single estimator dominated across all conditions, emphasizing that the choice of estimator should depend on the expected correlation structure and sampling design.