I try and fight them lies on both sides. And finally one ace reporter, William Freivogel of St. Louis Public Radio, sets the record straight regarding ProPublica's lie that that black teens are "21 times" more likely than white teens to be killed by police. This is the first light of day my lengthy bitching on the matter has received. Makes me almost think some of the time I waste trying to spread some truthiness is worth something.
But this isn't before that damn 21-times figure was repeated as fact by both a New York Times editorial and the Economist. [sigh] And what also bothers me is that I spoke at length with one of the ProPublica authors. He seemed to understand a) the sever statistical limitations of UCR homicide data, b) the statistical need for a larger "n," and c) the concept of a statistical outlier... but then he still refuses to publicly update or correct anything because the numbers -- statistically somewhere between meaningless and misleading -- are, well, computationally correct. I think he and his team are statistically savvy enough to know this is ideological bullshit.
One of the ProPublica authors recently doubled down: "We weren’t cherry picking years. We looked
at all of the years. But we were looking at what is happening in the
most recent years. The disparity is growing." But... but... they looked at the last three years. And if you look at just the last year or two (worth repeating: this is based on basically meaningless data), the number is actually shrinking.
It's one thing if they just happened to pick a year that was a statistical outlier. I could have pointed that out and they could say "Oops. my bad" (besides, it's not like the more correct numbers couldn't also help them make the same point). But to say they considered all the years and then chose the outlier that statistically means the least? Come on, now.
After reading this, take a break by watching David "1 of 3 Hanks" Klinger on the Daily Show. Klinger is the guy heavily quoted in that St. Louis Public Radio piece. Good stuff.