[ConsistentLibertarian]

"The typical Irishman is more prone to alcoholism than the typical Swede. This does not mean that all Irishmen or drunks or that Swedes are all teetotallers, but is a valid generalization based upon observation of group tendencies." OK, let's start with your example. You've got a way of distinguishing between two groups, and you've observed there's a statistical difference between them. Question: Does the criteria by which you seperated the two groups figure in the explanation of the statistical differences or not? If it does, then those criteria are playing a causal/explanatory role. If not then it's just a random connection. The criteria you used to initially divide the populations have no place in a scientific theory of the statistical data you've observed.