next up previous contents
Up: 2.3.2 Examples Previous: Approximate test with ``Extreme fishing'' (example of fishing for $ H_a$)

It is dangerous and may become meaningless if we choose the critical region $ C$ after we observed the data (or $ S$). Example: Suppose we have $ X \sim U[0,1]$ and we observed $ X = 0.62$. Set up test: In this case we will reject the null hypothesis $ 0.62 \in C$!

Eran Borenstein