Comment Re:Fingerprinting has never been scientifically va (Score 1) 103
The selection bias is only important if there's reason to believe that it has some effect on the item of interest. In this case, the dataset would be invalid if criminals have different fingerprints than non-criminals.
Besides, there are a LOT more fingerprints in databases than those of (reasonably) suspected criminals. More than enough to show there's no criminal-related bias. As an obvious example, perhaps you've noticed that the US is fingerprinting most foreigners who cross the border?
If all else fails you go to a university campus and get a bunch of non-criminals to volunteer their fingerprints. It wouldn't take an unreasonable number to show there's nothing special about the fingerprints of suspected criminals.