Greg Mayer writes: Have you seen this paper by Frank Corotto, recently posted to a university depository? It advocates a way of doing box plots using “comparative confidence intervals” based on Tukey’s HSD in lieu of traditional error bars. I … Continue reading …