Evaluation of different estimator approaches based on COVID-19 data and a simulation study

Abstract

Sampling is crucial in basic statistics due to the large size of certain populations. Sampling theory focuses on selecting samples based on appropriate population characteristics and approaches to estimating population characteristics from the sample. Firstly, we use the ratio-type estimator proposed by Prasad (1989). We adapt this unbiased ratio-type estimator to two new different estimators using the Hartley-Ross and non-response approaches. Finally, we combine these approaches and propose a new estimator. The theoretical results were supported by COVID-19 data and a simulation study with various scenarios. Based on the COVID-19 data result, it was concluded that the new tN3 estimator is the most effective estimator among those proposed and compared, with the minimum MSE and maximum PRE. Upon examining the changes in the COVID-19 data set in detail, it is evident that the PRE value increases with an increase in sample size, non-response rate, and k values. As in the COVID-19 dataset, the tN3 estimator was the most effective in the simulation study results. It is also important to note that these results are not applicable to only one scenario in simulation study. The tN2 estimator was the most effective in the scenario with rho=0.90, T=1000, n=400, non-response rate=0.30 and k equal to 2, 4 and 5.