Art Owen writes: Here’s a figure you might like. I generated a bunch of data sets constructed to have true effect size 1 and power equal to 0.06. The first 100 confidence intervals that exclude 0 are shown. Their endpoints … Continue reading