ubertuber

Is the question here whether arbitrary lines of social group or certain behaviors are the most powerful aggregator?

It seems like you are trying to parse whether people are a group because of some social cues or if they are a group because they shoot each other. Seems a little like missing the forest for the trees.

I know which classification would matter more to me if I was visiting.

The real weakness of host's point in my view is that the data set is small, specific to a few areas (which have more differences than crime statistics, such as enforcement practices, jobless rates, etc), and not really controlled too well.

__________________
Cogito ergo spud -- I think, therefore I yam