EFFICIENCY OF ROBUST ESTIMATORS AGAINST THE SAMPLE MEAN UNDER OUTLIER CONTAMINATION: A COMPARATIVE STUDY
Keywords:
Outliers, Population parameter mean, Sample mean, Robust estimators, MSEAbstract
In this study, the efficiency of some selected robust estimators such as the sample mean, sample median, 10%
trimmed mean and 10% winsorized mean are compared in the presence of outliers through simulation study. Data
used for this study was generated from R programming language under different distributions assumptions such as
normal, laplace and mixture of normal distribution with 10% identical outliers, in order to evaluate the robustness of
estimator under study. Mean square error (MSE) was used as a performance metric for each estimator. It was found
that the sample mean has the best performance in the presence of clean data (i.e. normal distribution). In the study of
outliers, it was found that the sample median has the best performance when data was generated from the Laplace
distribution as well as normal distribution with 10% identical outliers. In conclusion, the robust estimators are more
efficient and an appropriate alternative to the sample mean in the presence of outliers
