PERFORMANCE OF SOME PROCEDURES FOR DETECTION OF MANY OUTLIERS IN SAMPLES

Abstract: For detecting multiple outliers in normal samples we consider some tests and their performance and also depicted that which procedure is more powerful in the considered situation. The considered test statistics are ESD test, Grubb's test( Ek, LK and Lk*) and Rosner R-Statistics(RST). Here the power of these test statistics are studied and necessary discussion are made on the basis of computed result. Although ESD is an one outlier procedure we have used here to study the performance of many outlier procedure in case of detecting single outlier.

Keywords: Multiple outliers, ESD test, Grubb's test, RST, simulation, power.