One distribution, or two?
Was pointed to this thread by a friend of mine, felt I had to respond. Sorry about the length of the post.
Pure chance
could make this story. It's a classic case of a reporter being wowed by a standard distribution (previous discussion of teddy bears being apt), while not telling us what's important to know: one distribution, or two?
Code:
___X___
__XXX__
_XXXXX_
XXXXXXX
That's my best mock-up of a standard distribution. Now, we could have that, or we could have this:
Code:
___X______X___
__XXX____XXX__
_XXXXX__XXXXX_
XXXXXXXXXXXXXX
ie two distributions. In the first case, what you'd have is just the effects of chance. In the second, there's obviously something different between the two populations (it's rarely so clear, but there are ways to quantify it). Could be anything from when they go on patrol, where they go on patrol, their time on the force, or dead girlfriends in the basement. There are other tests to determine if there is a set of things that put a person in one category or the other.
So, it's not a case of a hoplophobic slant, a need for good copy, or a hatred of the LAPD, it's just simple innumeracy. This particular one is about as common as being thrown by regression to the mean. I imagine since the reporter has put in the effort on research, he'd be interested in finding out the truth. I've sent the paper an email about it, maybe we'll get a followup story?