Twitter | Search | |
Nate Silver 22 Nov 16
To follow: some *very* quick analysis which suggests the claim here of rigged results in Wisconsin is probably BS:
Reply Retweet Like
Nate Silver 22 Nov 16
Replying to @NateSilver538
Run a regression on Wisc. counties with >=50K people, and you find that Clinton improved more in counties with only paper ballots. HOWEVER:
Reply Retweet Like
Nate Silver
...the effect COMPLETELY DISAPPEARS once you control for race and education levels, the key factors in predicting vote shifts this year.
Reply Retweet Like More
The Notorious B.E.N. 22 Nov 16
Replying to @NateSilver538
also, how many likely clinton voters got turned away due to their new voting restrictions?
Reply Retweet Like
Para 22 Nov 16
Replying to @NateSilver538
I'm gonna pretend like I completely understand what's going on in that picture
Reply Retweet Like
Dan Hauge 22 Nov 16
Replying to @NateSilver538
Those demos would be the target of voter suppression laws as well.
Reply Retweet Like
Ed Vielmetti 22 Nov 16
Replying to @NateSilver538 @olmucky
You are crediting yourself with way too many significant digits in that analysis.
Reply Retweet Like
Tony 22 Nov 16
Replying to @NateSilver538
Why are you afraid of an audit? The fucking Russian gov't has bragged about interfering in our election!
Reply Retweet Like
Steven G. Harris 22 Nov 16
Replying to @NateSilver538
and an audit would certainly tell us whether your quick model was worth a damn this time
Reply Retweet Like
Timothy Verstynen 22 Nov 16
Replying to @NateSilver538
why does controlling for race/edu impact the paper/electronic disparity? Are those correlated in the data?
Reply Retweet Like
Nate Silver 22 Nov 16
Replying to @NateSilver538
Maybe a more complicated analysis would reveal something, but usually bad news when a finding can't survive a basic sanity check like this.
Reply Retweet Like
ffish 22 Nov 16
Replying to @NateSilver538
1) Why are you measuring the "shift" instead of predicting vote totals? 2) Why raw values instead of log(values)? [cont'd]
Reply Retweet Like
Gary Weissman 22 Nov 16
Replying to @NateSilver538
n=28 may be underpowered to detect the appropriate effect size; nice stata outpt but not convincing. All need better data.
Reply Retweet Like
ffish 22 Nov 16
Replying to @NateSilver538
Population isn't distributed linearly, and the four largest counties in WI use paper. Could skew analysis.
Reply Retweet Like
Lurker 22 Nov 16
Because paper and electronic ballots are used in different places, and different places have different race/edu
Reply Retweet Like
𝙰𝚕𝚎𝚡 𝙱𝚕𝚊𝚌𝚔𝚠𝚎𝚕𝚕 22 Nov 16
Replying to @NateSilver538
Clintonistas hope you're as wrong with this as you were on Election Day.
Reply Retweet Like
Jon Bender 22 Nov 16
Replying to @NateSilver538
Of course we don't know the full basis for Halderman et al's alleged claim, just thumbnail sketch in NY Mag.
Reply Retweet Like
Toby Pinder 22 Nov 16
Replying to @NateSilver538
feels very p-hacky yeah
Reply Retweet Like
J. Lang Wood 22 Nov 16
Replying to @NateSilver538
and yet you used the word 'probably' bs.
Reply Retweet Like
Patrick 22 Nov 16
Replying to @NateSilver538
thank you Nate. Appreciate you showing the regression!
Reply Retweet Like