The newest version of the conspiracy theory regarding Chrysler dealerships circulating the WingNet is pretty awesome. I just love it when they go old school:
This puzzled us. Why would there be a significant and highly positive correlation between dealer survival and Clinton donors? Granted, that P-Value (0.125) isn’t enough to reject the null hypothesis at 95% confidence intervals (our null hypothesis being that the effect is due to random chance), but a 12.5% chance of a Type I error in rejecting a null hypothesis (false rejection of a true hypothesis) is at least eyebrow raising. Most statistians would not call this a “find” as 95% confidence intervals are the gold standard for this sort of work. Nevertheless, it seems clear that something is going on here. Specifically, the somewhat low probability that the Clinton data showing higher survivability of Clinton donors could result just from pure chance. But why not better significance with any of the other variables? Why this stand out?
Then we got to thinking. Steven Rattner, the Car Czar, is married to Maureen White, one-time national finance chairman of the Democratic National Committee. What does Maureen do now? From her website:
Maureen White is currently Chairman of the Board of Overseers of The International Rescue Committee (IRC), a member of the North American Advisory Board for the London School of Economics, and a National Finance Chair of the Hillary Clinton for President Campaign. (emphasis ours)
That website looks dated, but you get the idea.
Someone wake me when they figure out how Monica was involved.
*** Update ***
I suppose part of what makes this so funny to me is I just got done reading a puff piece on Clinton, which essentially said that big bad Bill is sweet William now, and I thought that maybe the Clinton scares were over.
*** Update #2 ***
Someone else has noticed the seamless transition from the original hypothesis to the new and improved Clinton conspiracy. Unfortunately for the conspiracy theorists, that someone is Nate Silver:
In spite of this, Singer reports that “there [is] a significant and highly positive correlation between dealer survival and Clinton donors”. Although she hedges her conclusion a bit later on, this is a fairly irresponsible sentence to have written. Most people, in looking at this same exact set of data, would not only have avoided the implication that it proves the dealergate hypothesis, but would probably have come to something of the opposite conclusion: it argues strongly against the dealergate hypothesis. After all, there is no positive relationship whatsoever in the data on Democratic, Republican, Obama or McCain donations — which until Singer’s analysis was posted approximately 10 hours ago — had been the focus of the dealergate hypothesis. In fact, in several cases — such as for the data on Republican donations — the coefficient has the opposite sign of the one that the purveyors of the dealergate hypothesis were hoping for. Republican donors were incrementally less, rather than more likely likely to have their dealerships shuttered, according to Singer’s analysis, although the pattern is nowhere in the ballpark of being “statistically significant” as most of us would define it.
Predictably, this has not prevented people like Michelle Malkin and Doug Ross from claiming that Singer’s data confirms their hypothesis. Of course, it does not confirm their original hypothesis, which was that donors to Republican candidates were more likely to have their dealership closed. Instead, a new hypothesis has evolved — it’s all about those dirty, rotten Clintons! — the sole reed of evidence for which is Singer’s overstated conclusion (but not really her underlying data itself).
***Why, after all, stop at Clinton donors, who until this morning had never been central to the dealergate hypothesis? Why not look at John Edwards donors, or Ron Paul donors, or donations to any of various political action committees, or donations to members of the Senate Banking Committee, or donations to Congressmen who voted for the auto bailout plan? If you looked at enough of these, you would eventually come up with a few positive results — and then you could work backward to formulate your own conspiracy theory around it. There is a name for this sort of practice: data dredging.
At the end of the day, people are going to believe what they want to believe: some people believe that the moon landing was faked, that 9/11 was a grand conspiracy, and that Barack Obama was born in Indonesia. There is no evidence for any of these claims, but that doesn’t stop tens of millions of people from believing them! Dealergate, particularly in its original formulation (that Obama was punishing Republican donors with the Chrysler closings), is in largely the same category.
These folks are really just hopeless.
MattF
“Nevertheless, it seems clear that something is going on here.” A bit of Statistics for Dummies: if you ever find exactly 50-50 probability for some event– it’s not random.
dmsilev
Wrong Clinton conspiracy. The real question to ask was “Did Vince Foster drive a Chrysler, and why was he *really* killed?”
-dms
Calming Influence
They’re subconsciously thinking of hummers…
bago
Moar reverse vampires!
John D.
Dear God, I haven’t seen data mining that aggressive since Statistics 301. Pure masturbatory statistics play.
“Maybe” is the best answer he/she can come up with? How about “maybe” they went looking for their preferred conclusion.
Idiots.
The Grand Panjandrum
That piece is going to take some third year math major about 5 minutes to refute. I’m just to lazy to even bother with it. Jesus! A very simple back of the envelope calculation (which Nate Silver has already done, and you linked to previously) will show that these mouth breathers got nothing.
Martin
Look, there’s a 50/50 chance that this is retaliation by Clinton/Obama. Either these dealers were going to go public with their material support for covering up Vince Fosters murder and planting supporting evidence that shows that Obama isn’t a muslim terrorist from Kenya, or they weren’t. That’s 50/50, and so I don’t see how we can discount a 50% probability that this was a means to shut up the people about to reveal 20 years of conspiracy.
Ash Can
Manufacturing tie-ins with Bill Clinton is the right-wing sudoku.
Joshua
Have you guys ever seen the dreadful cryptozoology show on History Channel, MonsterQuest?
This is about on that level of scientific investigation. And the conclusion is even presented the same way! “Well, we can’t prove that the Ohio Grassman/a massive Clintonian conspiracy revolving around campaign donations and Chrysler dealerships exists… but if you kinda squint a bit at the data, you can’t prove that it doesn’t! It would be irresponsible not to speculate that the Ohio Grassman is going around bankrupting Chrysler dealerships owned by Republican donors!”
DonBoy
You’re not getting the full intellectual sin here. Their preferred finding is that Obama’s rewarding people who gave to him in the primary, against Clinton. But, having found the exact opposite, they need to able to show how that’s just as bad.
Little Dreamer
It was there in the Bible all along!
Jay C
Jeez, guys: this is really the way to make your “point” –
Does this tendentious mathbabble actually mean anything, or is this merely meant to sound “authoritative” – like the wordy and elaborately argued “proofs” that spoofers occasionally crank out to “prove” that 2+2=5?
shirt
Hasn’t wingnuttia not realized the correlation between republican attitudes and used car dealers? Don’t they both approach the common man with the same intent as a snake does a mouse?
Joshua
@Jay C: Actually, what they’re saying is, “Our analysis found absolutely nothing, and we’ll admit that. But we’ll go on anyway pretending like we did.” The math talk is just to distract people by making it sound like they’re doing some kind of serious analysis. They aren’t, they’re just aping what scientists do without actually understanding it.
Steeplejack
Dude, be careful using the phrase “puff piece” in connection with Clinton. Just sayin’.
burnspbesq
Even if it were true, it would be insufficient payback for all the years that the incidence of AMT has been heavily skewed toward blue states.
RSA
Amen, Jay C. Jargon an expert doth not make. I liked this bit the best:
Using this brand of statistics, a p-value of 0.125 is not significant in the first place. So, to answer the question “Why would there be…?” the answer is “There wouldn’t.”
vishnu schizt
Well what all of you dirty fucking hippies forget is that you all use “scientific analysis” to convince yourselves that Saddam had no WMD’s, and that the earth isn’t 6000 years old. So why do you get your goddamn patchouli in a lather when we prove “scientifically” that the probability exists that Clinton shut down car dealers as a vendeta (or how ever the fuck you spell it) for not contributing to a losing campaign. I’m sure your messaih wants to keep the peace with Clinton, and offered this up so that the “whitey tape” wouldn’t be released. All of you libtards are so smug, think you are so smart, while you eat your fancy cheese, drink wine and drive you fancy car! Fuck you assholes, I like my cheese processed, my beer brewed with rice and my shit hole chevy on blocks. I fully understand why that cornholer Sullivan hates you dems.
trizzlor
@RSA: Exactly, anyone who thinks a p-value of 0.125 is significant but not significant aught to be banned from math. What’s worse is this update:
So they sliced and diced the data faster than you could say Bonferroni and even that still gave them nothing.
Quiddity
Wikipedia’s entry for p-value contains an example for value=.115
It’s not a statistic to get excited about.
matoko_chan
Please….a linear regression? Too crude.
This validates a hypothesis I am formulating.
I had thought party affiliation stratified by IQ, but it appears that conservatism actually kills brain cells!
And conservatism is apparently deadly to cool cells too.
myiq2xu
Irrational Clinton hating never goes out of style.
toolshed
This is the kind of stat analysis you get from the frat douche bags in my intro stats class who just try to cram all the jargon he can remember into their final papers praying I just just end up skimming them for key words.
In reviewing that for less than a minute he seems to make the following mistakes:
*He uses several highly correlated explanatory variables in his model which is going to completely fuck his standard errors up (meaning his p-values are probably massively understated)
*He then apparently got preliminary results and just stopped, regardless of the fact that none of his variables were significant. That “12.5% type 1 error” rate is complete bullshit (regardless of the fact that he obviously doesn’t really understand what that statement means). The bast conclusion he could really make is th\o say that “if I create this model and randomly put in shit to take out a bunch of the response noise I still can’t prove a goddamn thing significantly”.
I would say more but I have already put in more effort than I would grading a term project that read like this. If a student handed this in I’d just put a giant x through it and write “nice effort but how about you pay attention when you retake this class next time”.
toolshed
So, for some reason I couldn’t stop thinking about this.
You have nothing but categorical variables and a categorical response. There is no excuse to be using linear regression on this problem… unless that is its the only tool you know, which I think is a safe assumption on his part. Thats actually a good thing really because then he can plead ignorance rather than malice on this.
If he doesn’t know how to run an anova model or any other more reasonable approach to this analysis, then there is no reason to believe he can properly run a regression on the log-odds which I believe is what he treid to do. So even these results are suspect, which is in addition to my criticisms above on variable choice.
Lastly, if he’s willing to accept the other variables to be insignificant than he has a clinton indicator vs a dealership closing indicator. That’s a 2 sample z test for proportions you fucking moron! I’ve seen some dumb motherfuckers come through my classes who were even able to recognize that.
If he had done that test maybe I would have given him a pass,even though the z test is just an approximation, but anyone who knew what they were talking about would have used the Fisher Exact Test in this situation.
I’m an idiot for getting so worked up about this shit but people lying with statistics ruins the credibility of the whole field.
trizzlor
@toolshed: I think the reason they didn’t run an ANOVA, aside from not knowing how to, is that their point wasn’t to explain the bulk of the variance in the model but specifically to tease out any kind of Dem/Closed correlation. I can see no other reason for running that many independent tests, and on top of it all leaving it to “the readers to judge the import of this observation” without doing even a rudimentary multiple test correction.
I agree that Fisher’s Exact or a Chi-Square (if they have enough samples) is the obvious choice and immediately jumping to a regression reveals ignorance or bias. If I saw this question on an exam I’d assume it was one of those initial throw-aways to make sure the student is alive.
trizzlor
argh, I can’t get enough of this. This comment is priceless:
Rather than give it up or use a simpler analysis of variance, they’re actually introducing even more biases and constructing “scenariables”. I hope no one tells them about Weka, they’ll be up all night making decision trees.
Perry Como
I lold.
Jay Levitt
You statistics buffs are using way too many words, and it turns this into a one-way hash argument. I can do it in three: Texas sharpshooter fallacy.
You can take any data set and find SOME correlation with SOMETHING. Hence the global warming/pirates connection, Bible “codes”, etc.
Brachiator
@The Grand Panjandrum:
That 16 year old kid who recently solved a centuries old math puzzle would look at this wingnut stuff and say, “this is some lame BS!”
trizzlor
Malkin just hit this, prepare to see the same argument made by a Republican congressman next week:
John Cole
@trizzlor: That is pretty funny.
Nylund
Let me step in to tackle this quote:
I just wrote a long and boring response to this and erased it. But basically, she obviously doesn’t understand the concept of a type 1 error. Rather than explain statistics, I will use an analogy.
Pretend you think you might be pregnant. You go to the doctor. The doctor tests you and says, “you are not pregnant.” Then, she comes busting in the door and says, “Yeah, but even if the doctor said you were pregnant, there is a 12% chance that you really weren’t!”
For some reason, she thinks that telling you this will make you more likely to conclude that you may in fact actually be pregnant. I have no idea why she thinks this. I suspect she doesn’t know a thing about statistics, read about it online for about 5 minutes, and got really confused.
Considering that she can’t even spell statistician, I think this is a fairly reasonable guess.
bago
12.5% is 1/8.
Nylund
I thought I’d amend my earlier comment with an explanation of what I think she was trying to argue. I’ll use another analogy:
In law, we assume that people are innocent until proven guilty. The DA must come up with evidence beyond a reasonable doubt in order to convict someone. This makes it very hard to convict people. Theoretically, this means that a number of criminals probably go free, but very few innocent people end up in jail. This is how we do thinks in statistics as well.
Alternatively, we could assume that everyone on trial is guilty, and require people to prove, without a doubt, that they are actually innocent. This would likely result in very few criminals getting away with their crimes, but it would also mean that a lot of innocent people might end up in jail because they couldn’t prove their innocence.
She is saying that when you assume that Obama is innocent, you can’t prove he is guilty. I think she is then attempting to claim that if you did things the second way, and assumed he was guilty, well then it might be hard to prove he was innocent.
But, in the real world, we don’t waste our time trying to prove every crackpot theory wrong (should we fly to Saturn to prove its not made of play-doh?). No. We require crackpots to prove they are right.
Gordon, The Big Express Engine
@Nylund: That’s funny about the crackpots. A co-worker of mine recently stated that “Reagan reduced the size of government by a third and the economy grew as a result.” I called bullshit on that asked him to show me proof. He then said I had to show proof that he was wrong. I said that he was the one making the ridiculous claim and he had to back it up. Needless to say, I am still waiting..,
Brachiator
@trizzlor:
Ergo, Hillary Clinton killed Vince Foster.
New rule: As wingnut arguments intensify, the odds that Hillary Clinton will be blamed for something approaches certainty.
Perry Como
Fixt for wingnuts.
Left Coast Tom
If the Obama administration hadn’t worked on a Chrysler rescue there would now be _zero_ Chrysler dealers remaining. Now we’re arguing over how many of the survivors, who wouldn’t have survived otherwise, were donors to which candidates/causes? Wow.
k55f
And coincidences are just coincidences.