We first consider Basu's Theorem and a number of applications of it. In particular, we show how it can be used in distribution-free situations. This allows us to select an appropriate nonparametric test after looking at the data in a certain way and still maintain a given significance level. That is, we can legalize what some statisticians call "cheating by looking at the data before selecting a procedure." This suggests that we use robust methods that are based on good selector statistics. We end by showing how most "robustniks" handle data so that outliers do not have too much influence and how these procedures can be changed (adapted) with an appropriate selector statistic.