To Remove or not to Remove: the Impact of Outlier Handling on Significance Testing in Testosterone Data