New robust class of estimators for population mean under different sampling designs

Abstract

New robust class of estimators that auxiliary variable information such as coefficient of variation, kurtosis and quarters are used in are suggested by considering robust methods of the population mean under different sampling designs. The estimators that using LTS, Huber MM, LMS, Tukey-M, LAD and Hampel M robust methods are investigated under simple random sampling (SRS), ranked set sampling (RSS) and median ranked set sampling (MRSS) designs. A real data example and detail simulation study are applied to see efficiency of proposed robust class of estimators in SRS, RSS and MRSS designs. The efficiency of suggested estimators are compared with regard to mean squared error (MSE) and percent relative efficiency (PRE). The numerical study is used to observe performances of the estimators concerning body mass index (BMI) dataset. In the simulation study we considered seven continuous probability distributions, five sample sizes with different sampling designs. In general, our results show that the proposed robust class of estimators performs better under MRSS.