Generally, power increases with sample size. But if you look carefully at the plots on page 64, you will see that although there is an overall increasing pattern to the plots, power is not quite monotone in the sample size. (You may want to recreate the plots at a larger size to see this more clearly.) That is, sometimes increasing the sample size actually decreases the power. This seems counterintuitive. Can you give a plausible explanation for why this is happening?