I get that some folks thought we were being too hard on Silver over the Pielke stuff. Fine. But look at this recent piece of utter dreck. Anyone with a basic understanding of statistics (ie not Benjamin Morris) would know that you don’t suss out meaning in regressions using highly highly correlated numbers.
Multicollinearity is an actual thing. You have to consider interaction terms and It’s not a difficult thing for actual quantitative researchers to know. And you know what? A basic understanding of demographic and nutritional data would know that weight and height are HIGHLY correlated variables, especially in such a selectively created population like the NFL. Otherwise the result is just numerology. May as well hire Dan Brown to write articles about quarterbacks, seriously.
Wait, he compare t-values (which he put in quotes) to determine which was a “better predictor”?Report
Yeah I don’t get it. Why not just show the step-wise R^2? Then we’d also have an idea of how meaningful the model is overall.Report
From what I can tell, there were only 3 parameters in his model of three sets of incredibly complex numbers (that is, there is a ton of highly variable stuff that goes into producing those numbers). I can’t imagine the r2’s would have been very high, or telling. This sort of model is perfectly OK if you’re just trying to get a sense of something before you delve into it deeper. I wouldn’t write an article about it, though. I’d just keep it to myself when I built a better one.Report
Which players have heights that are complex numbers?Report
I don’t think Chris is using “complex number” in a technical sense.Report
it’s a different sport, but Fletch’s height was partly imaginary.Report
I’m disappointed with fivethirtyeight thus far. Little in it holds up to the excellent writing in Silver’s book or his writing at the NYT; and much feels so factoid and separate from human stuff as to be bone dry dry dry.
But Ezra Kline’s Vox has been a joy to watch. I’m particularly fond of the whole ‘card’ notion — that you can keep building on a story in this way, making the older info readily available so that the whole thing is part of a whole, is a beautiful thing.Report
If it was even factoid based I’d be less critical. This isn’t factoids or dryness, this is INNUMERATE factoids.
Nate Silver promised us Bill James crossed with Edward R. Murrow. He gave us Skip Bayless crossed with Dan Brown numerology.Report
He was drafted by the Browns. There are few better predictors of future failure (drafted by the Raiders being one).Report
This is where I would ordinarily make a Sooner joke, but the fact that there wasn’t a single Longhorn drafted this year makes my ground much too shaky.Report
Gee @nob-akimoto I usually love your contributions here, but your true colors are showing. (really who looks good in orange?!)
Boomer!Report
In baseball, it would be “A free agent signed to a long-term contract by Brian Sabean.”Report
Man, that’s terrible.Report