[jagthebeetle]

Agreed, not least because: - area-based visualizations make the effect hard to distinguish; bar charts or data clouds with numbers and confidence intervals would have been way more immediate. - the colors make the negative group (usually) more visually prominent, since it has higher contrast with the background, exacerbating the area-estimation problem. (e.g. me wondering, "are there more overweight pink people as a fraction of pink people?")