Thursday, April 2, 2015

Lies, Damned Lies, and Statistics, Again

Reading a few articles over on, I'm overwhelmed by the logical fallacies and failures of reasoning that the readers are supposed to just swallow without question. Permit me to "mansplain" one of the more problematic ones.

Let's say we have ten targets, and we line them up left to right along a wall. Standing just a few feet away, I take a bow and arrow, aim at the target furthest to the right, shoot, and hit that one target. Ten percent of the targets on that wall have now been hit by an arrow.

This does NOT mean, however, that, "If you're a target on that wall, you have a ten percent chance of being hit by an arrow." It means that the one I aimed at had a very likely chance of getting hit (the exact percentage depends on such variables as my marksmanship and the quality of the archery equipment), while the ones I wasn't aiming at were very unlikely to get hit at all. The chances that my arrow would have left the bow, taken a sharp left turn, zoomed to the target furthest to the left, taken a sharp right turn, and then planted itself in that far-left target are something close to zero percent.

And yet that's exactly the sort of assertion being made when  they say that a person of color has a 44%-50% chance of going to prison or that a woman has a one-in-three chance of being raped. It presumes that there is a sole, relevant variable--in these examples, being a person of color and being a woman, respectively. For that to be true, we have to assume that because President Obama is a person of color, he has the same chance of going to prison as does a black gang member who sells crack in front of a police station, or that an old woman alone on a sailboat in the middle of the ocean is as likely to be raped as is a teenage girl in sub-Saharan Africa who's been kidnapped by enemy soldiers. That kind of claim doesn't pass the most basic of credibility tests.

No comments:

Post a Comment