DETECTION OF MULTIPLE UPPER OUTLIERS IN EXPONENTIAL SAMPLE UNDER SLIPPAGE ALTERNATIVE

Abstract: This paper aims to compare the empirical powers of some statistics for detecting multiple upper outliers in exponential samples under slippage alternative. In addition to that we also investigate masking effect for various degree of discordancy parameter. The results which are based on simulation study, indicate that the maximum likelihood ratio test statistic is better than the other statistics followed by Dixon type test statistics to deal with upper outliers in exponential samples. Also in case of combating masking the maximum likelihood ratio test statistic is better and the test proposed by Lalitha and Kumar (2012) precede Dixon type test.

Keywords: Outlier, slippage alternatives, masking effect, discordancy parameter.