[NOTE FROM JASON, 10/25/17 12:14 EDT -- I'm opening up the Blogger comments at the bottom of this post so that there can be a linear record of this conversation rather than Twitter threading, which is hard to follow... don't make me regret it, kids!]
-----------------------------------------------------------------------------------------------------------------
A year ago, leading up to the election, one of my biggest worries wasn't yet what Comey's influence on the election was going to be, but rather what the impact of voter suppression would be.
On election night, watching the results come in, I sat at a volunteer desk at Hillary's Brooklyn HQ and scanned county-level maps in MI and WI long before the states were called for Trump. MI was going to be tight because the turnout in the Western suburbs of Detroit (like Dearborn and Ann Arbor) was far lighter than it had to be. WI, on the other hand, was lost. I looked at the numbers coming in from Milwaukee and Madison versus the rest of the state and saw what would turn out to be an insurmountable Trump advantage. The fringe Obama voters hadn't turned out... Or had they?
Days after the election theories were flying about faulty voting machines, Russian hacks, long lines of people being turned back, and people being denied for not satisfying the overtly racist Voter ID laws. The first two were an easy out: it wasn't our fault. America knew better than Trump but Russia stopped her from electing Hillary Clinton. I still believe that there was some illegitimate vote tampering that will be discovered as the evidence continues to roll in. But some took it on their own initiative to examine precinct-level results in Wisconsin and found a correlation between precincts that used voting machines and precincts that went for Trump. Fivethirtyeight looked into it and so did Nate Cohn:
Along came Mike Farb (@mikefarb1) who promised his own independent investigation. I, along with many others, followed him and waited to see what he would come up with. His initial results, while captivating, were really just a rehash of the New York magazine piece which had already been disproven. Curious, I asked him for his methodology:
It soon became apparent that he was not looking for help, rather he wanted to continue asking for donations to support his dubious results.
As a statistician, it is abundantly clear to me that Mike Farb wants no part of an actual statistical methodology because that would show his results are junk. And it's exceedingly tough to grift people when you don't even have the appearance of something to offer them.
If you want to continue reading him, that's great. We could all use a little bit of fiction. But please don't feed the grifters.
UPDATE: Looks like I'm not the only statistician he's blocked.
-----------------------------------------------------------------------------------------------------------------
A year ago, leading up to the election, one of my biggest worries wasn't yet what Comey's influence on the election was going to be, but rather what the impact of voter suppression would be.
On election night, watching the results come in, I sat at a volunteer desk at Hillary's Brooklyn HQ and scanned county-level maps in MI and WI long before the states were called for Trump. MI was going to be tight because the turnout in the Western suburbs of Detroit (like Dearborn and Ann Arbor) was far lighter than it had to be. WI, on the other hand, was lost. I looked at the numbers coming in from Milwaukee and Madison versus the rest of the state and saw what would turn out to be an insurmountable Trump advantage. The fringe Obama voters hadn't turned out... Or had they?
Days after the election theories were flying about faulty voting machines, Russian hacks, long lines of people being turned back, and people being denied for not satisfying the overtly racist Voter ID laws. The first two were an easy out: it wasn't our fault. America knew better than Trump but Russia stopped her from electing Hillary Clinton. I still believe that there was some illegitimate vote tampering that will be discovered as the evidence continues to roll in. But some took it on their own initiative to examine precinct-level results in Wisconsin and found a correlation between precincts that used voting machines and precincts that went for Trump. Fivethirtyeight looked into it and so did Nate Cohn:
I wanted to believe that there was some merit to their findings but I replicated the results that the two Nates each independently put out and came to the same conclusions. When you factor in demographics, the impact of voting machines comes out to be statistically insignificant.These guys might know about computer security, but they probably don't know anything about elections https://t.co/5khbcVDyhr pic.twitter.com/i42duBpnw0— Nate Cohn (@Nate_Cohn) November 22, 2016
Along came Mike Farb (@mikefarb1) who promised his own independent investigation. I, along with many others, followed him and waited to see what he would come up with. His initial results, while captivating, were really just a rehash of the New York magazine piece which had already been disproven. Curious, I asked him for his methodology:
What was your methodology? @Nate_Cohn dumped on a prior study for cloudy methodology. If you want to draw statistical conclusions, /1— Daily Comb Over (@tgcCombover) May 10, 2017
you need to be sure to control for fixed effects across the individual precincts. A difference in differences estimation is an option /2— Daily Comb Over (@tgcCombover) May 10, 2017
as well. But you'll have to be sure the treatment precincts have comparable controls. Basically, did the inclusion of EVCs increase Trump /3— Daily Comb Over (@tgcCombover) May 10, 2017
Vote moreso than the typical trend shift you see in the precincts without EVCs. Your best bet is to compare precincts within a county /4— Daily Comb Over (@tgcCombover) May 10, 2017
for the cleanest comparisons. My hunch is you're correct but let's be sure no questions can be asked about methodological approach. 5/5— Daily Comb Over (@tgcCombover) May 10, 2017
It soon became apparent that he was not looking for help, rather he wanted to continue asking for donations to support his dubious results.
He continued to ignore my requests and then went on to Tweet about how Southern States had been experiencing stolen votes since the Nineties. Basically he was ignoring the party realignment that had occurred. Instead, he was chalking up the Republicans' statewide election wins to stolen votes. Of course, anyone who did in minute bit of research about politics in the South would know that while registered Democrats outnumbered registered Republicans, many white Democrats were just residual dixiecrats who had not yet switched party affiliations, but had become reliable Republican votes since 1994.Word of caution: his methodology is flawed and he's running with a very much unproven conclusion.— Daily Comb Over (@tgcCombover) May 13, 2017
It's interesting but probably not true
No. There are still voters registered as Dems in the South that never switched over after 1994 despite never voting Dem again.— Daily Comb Over (@tgcCombover) July 13, 2017
Each week he put out more graphs and charts without any formal methodology. Finally, I called him out on it. I ask him for his methodology once again and one of his associates pointed me to his website. I outlined three major points that were wrong with his work, and offered to help:Why not focus on real problems instead of making stuff up? There are lots of legacy Dem registrations in the South left from 94. Meaningless— Daily Comb Over (@tgcCombover) July 13, 2017
But as I pointed out to you several months ago, your proof in WI didn't involve statistical controls so you really cannot draw causality.
— Daily Comb Over (@tgcCombover) July 31, 2017
I want to believe your story because it would have an easy solution. But I just don't trust your improperly defined statistical analyses.— Daily Comb Over (@tgcCombover) July 31, 2017
Switch to paper ballots. That obviously is a better method than machines, but don't expect it to be the end all. No evidence for that yet.— Daily Comb Over (@tgcCombover) July 31, 2017
In a district with a large number of old white racist voters who viewed the Scalise shooting as the start of a war against them. Yes we agre— Daily Comb Over (@tgcCombover) July 31, 2017
This is going back to your WI analysis. You promised me you'd share it then and I'm still waiting. Instead you insult me and call me a Foxer— Daily Comb Over (@tgcCombover) July 31, 2017
Good luck. As I said, I hope it is that simple.— Daily Comb Over (@tgcCombover) July 31, 2017
I just want to see proof.
If someone is misleading people by drawing conclusions that aren't proven out of "simple graphs," and then asking for donations that's wrong— Daily Comb Over (@tgcCombover) July 31, 2017
It is cumulative votes precinct shares analysis. The methodology is in any post grad stats curriculum. https://t.co/rZaMb7KxSx— 3M-TripleM (@SwissTriple_M) July 31, 2017
The methodology 4 the graphs produced by SAILL is not flawed. I have compared her results to my results &my results were checked by 2 PhDs.— 3M-TripleM (@SwissTriple_M) July 31, 2017
It's on YouTube: https://t.co/Rd9dWkybtN an Saill's website CLEARLY explains the methodology used. No secrets there: https://t.co/5Fg6MGtSES— 3M-TripleM (@SwissTriple_M) July 31, 2017
Send me your methodology.— Daily Comb Over (@tgcCombover) July 31, 2017
I found similar results to @Nate_Cohn.
I really want to believe this is causal but the numbers don't suggest it. pic.twitter.com/yVug4L9izY
My questions to you (And they will come in slowly):— Daily Comb Over (@tgcCombover) July 31, 2017
1. How can you insist there is usually no demographic relationship politically?
2. You mention precinct data but your graphs report county results. Precincts results can definitely vary within each county, right?— Daily Comb Over (@tgcCombover) July 31, 2017
3. You attribute that to precinct size but have you actually looked at demographics within those Precincts? See Q1.— Daily Comb Over (@tgcCombover) July 31, 2017
Thanks again for sharing. And despite being insulted by @MikeFarb1, I've enjoyed the discussion. As I said, I really do hope to help if poss— Daily Comb Over (@tgcCombover) July 31, 2017
We don't insist. We look for demographic/precinct size relationships. Sometimes we find them, often we can't.— Saill (@saill) July 31, 2017
Right. So what you're finding are correlations. For causality you need to show that precincts with higher vote totals were falsely swayed.— Daily Comb Over (@tgcCombover) July 31, 2017
That variation can potentially blow a hole in your results! pic.twitter.com/j1ioHzd8Zw— Daily Comb Over (@tgcCombover) July 31, 2017
But again, a separate issue. Machines are hackable. That doesn't mean they were hacked. It means they may have been. See the difference?— Daily Comb Over (@tgcCombover) July 31, 2017
As I said, I'm glad to help if I can. 🤔— Daily Comb Over (@tgcCombover) July 31, 2017
You guys are just drawing too big of a conclusion out of a narrow finding. That's my critique.
That's why we aggregate county level results for each state. One would expect blue shifts to cancel red shifts, would one not?— Saill (@saill) July 31, 2017
Yes, it depends on relative magnitudes. They might not completely exactly cancel out. But why would swing states look different from others?— Saill (@saill) July 31, 2017
Demographic similarities in upper Midwest, rust belt, etc.— Daily Comb Over (@tgcCombover) July 31, 2017
How do these compare with same precinct results in 2012?
In various ways, yes.— Saill (@saill) July 31, 2017
And that means that in order for Democrats to win, we need more votes in blue precincts then they get in red precincts.— Daily Comb Over (@tgcCombover) July 31, 2017
All of the actual evidence shows that voter suppression plays a role in preventing that from happening.— Daily Comb Over (@tgcCombover) July 31, 2017
So while switching to paper ballots is clearly preferred, statistically speaking it is unlikely to solve the problems you are showing— Daily Comb Over (@tgcCombover) July 31, 2017
For a full account (since I was blocked shortly there after and his replies are unavailable to me), just do a twitter search for "@mikefarb1 @tgccombover."Best of luck with your work moving forward.— Daily Comb Over (@tgcCombover) July 31, 2017
As a statistician, it is abundantly clear to me that Mike Farb wants no part of an actual statistical methodology because that would show his results are junk. And it's exceedingly tough to grift people when you don't even have the appearance of something to offer them.
If you want to continue reading him, that's great. We could all use a little bit of fiction. But please don't feed the grifters.
UPDATE: Looks like I'm not the only statistician he's blocked.
Same here. Asked polite questions; he blocked. Statisticians don’t do that to statisticians, like ever: we critique each other & share info.— SECRETARY NETHERY (@shastabarbell) October 25, 2017
No comments:
Post a Comment