Many years back, when Mark Germain was known as “Mr. KFI,” a doting caller told him he was “one in a million.” Germain replied that that means there are 7 people in the greater Los Angeles area who are exactly like him. In that vein, Jason Felch and Maura Dolan of the L.A. Times have an interesting article on “Arizona searches,” whereby a statewide DNA database is searched for internal matches to see how many individuals within it randomly match to each other. Most of these matches are to be expected. After all, while the odds of two DNA profiles matching are extremely remote, the odds of someone matching randomly to someone else are damned close to one if you repeat that experiment enough times. In the case of Arizona’s database, which included 65,000 unique records in 2001, that means Kathryn Troyer effectively made 4,225,000,000 comparisons in one fell swoop, yielding 100 expected matches and 44 unexpected ones. The article doesn’t say what the statistical margin of error was, so it’s impossible to tell from the article - and perhaps from the Arizona study alone - whether the usually cited odds are accurate. The real issue here, as far as I am concerned, is that the FBI is asking us to trust their estimates, while fighting tooth and nail against any serious efforts to test them empirically.

Also chuckleworthy is the objection that testing an entire database against itself greatly increases the odds of a match being found. Of course it does, but so too (albeit to a lesser extent) do database searches, whereby a single DNA sample is tested against an entire database of non-suspects as a method of choosing a suspect (cf. Mike Nifong’s infamous “no wrong answers” police lineup). If we are serious about DNA evidence as science, rather than AGW-esque “consensus” that we are all just bullied into believing without proof, then Arizona searches ought to be conducted nationwide. Then the results should be published and explained, in context. If the tests ultimately show that the FBI’s best estimates were correct, great. But if they show anything else, we need to know.

UPDATE: Instapundit has more. Patterico, alas, has less. David Kaye has much more.