The Impact of Outliers on the Power of the Randomization Test for Two Independent Groups

In our work we studied the impact of outliers on the power of the randomization tests for two independent groups. Four variables were manipulated: simple size, number of groups with outliers, type of distributions and effect size. To simulate data with outliers Normal contaminated distributions were used. For the simulations, we used R software, namely the package [rsquo;nor1mix]. We observed that randomization test loses power for all distributions with outliers when compared with the power achieved with the Gaussian. Our results were illustrated graphically and interpreted. We observed that loses in power are directly related to the magnitude of the outliers and are inversely related to sample size.