I would like more data for [1], with a breakdown per gender of each category and size. The data they show could be under the effect of a Simpsons paradox from the way they present it.