It seems likely that the number of observations is either of interest in it’s own right, in which case it should be plotted separately, or it is used to indicate the confidence in the resulting mean adoption rate. If the latter, then a better approach is to plot the mean point with error bars giving the 95% confidence interval on the mean. Large numbers of observations will result in narrower confidence intervals. This would also eliminate potential confusions caused by having a point that extends, like a continuous variable, along the x-axis. ]]>

A.

]]>I’ve heard warning about such plots as well: you don’t know if the reader is attending to the diameter, area, or volume. Hopefully, our quiz will give some insight into how tough it is for people to eyeball areas.

]]>