Pages

Sunday, November 13, 2016

Lies, damned lies, & statistics

Short term update:  Tweeted here by the great Nassim Taleb.  Polling statistics from Statistical Ideas topped 4 million reads in the past month and >5 million cumulatively.  Latest election articles are all searchable here. As a reminder, being among our 150 thousand Statistical Ideas followers is easy to do, through e-mail@salilstatisticsfacebookLinkedIn, Instagram.



Let's review why 20th century lecturer, Mark Twain, would have expressed the following as a parting thought on this past week's presidential campaign:
There are three kinds of lies: lies, damned lies, and statistics.


Lies
The site has consistently bore condemnation for way too long, just for inoffensively reproaching “professional polls” as bogus. But now, for being correct, a lot of hard-earned reverence from an even bigger audience.  Here is a summary of the nicely forewarning, multi-page research articles we published, in just the past couple months, and that collectively amassed over 3 million reads and thousands of shares, including from senior advisors and surrogates on both sides of the aisle.  For perspective on that number, it's equal to the print subscription/circulation of my cherished The New York Times and Wall Street Journal combined (and a typical mainstream media article attains only several thousand reads).

Presidential coin flip
Presidential betting markets 6x money
Antagonism isn't perpetual
Sea of faulty polls
Pollsters Gone Wild
Abstruse pollsters

We should note that last week's popular vote came down to ½% (~½ million people). So if you and your network can reach 3 million people and flip just 5% of them, then you are almost there in moving the needle!  And sad to say it, but the popular vote is now just a face-saving falsehood at this point by the very ensemble of “highly competent leaders” stunned this all happened: we often see immaterial popular vote results by random chance, when neither candidate is absorbed by it. Please don't degrade yourself by abruptly rioting on major U.S. streets, as a response to not winning a fair election!  As a statistics professor at a few top universities, I found being instructed to cater to election-result anxiety as an unprecedented excuse.  Donald Trump's forceful lead in the electoral vote would have had a >2/3 chance of spreading straight across to the popular vote, if that were ever our main objective.



Damned lies
Now reactively impugning FBI Director, James Comey, as well is simply that, the latest in an endless unfamiliarity of rational data, which anyway most Americans are rightfully repulsed by.  Instead of blaming e-mail mendacities by Hillary Clinton or Huma Abedin or Carlos Danger, or exploiting Trump's video leak, or the alienation of anyone who got in the way (including Bernie Sanders' supporters), it just seemed cozier to blame Russia, Julian Assange, and hate.  But the truth is her delicate spread had started crashing days prior to the Mr. Comey re-enquiry statement (one she falsely claimed only went to Republicans).


And statistics
Obviously detectable a mile in advance for anyone who cared: major "pollsters" and models, each with an advertised "margin of error" of about 5% on their Hillary Clinton odds.  Yet all were collectively corrupt, and incoherently mistaken in the same direction. Pollsters so bad, they make fortune tellers look good.  We oft-times noted, those in the 90's% at any time during 2016 (which they all were), were always wrong in everything from demographic selection, to understanding what's a probability calculation.



It's always a pleasurable day when the luminary Nassim Taleb (NYU professor, black swan author, and follower) enjoys our particularly heartless assessment of foremost pollsters.  And given elections just now, splendid that this is his (and my) top tweet in the past couple days!


As a particular foundation for much of the misrepresentation, comic Nate Silver was wholly wrong this entire election cycle.  First putting Mr. Trump at 2% during the primaries, and then having him crazily oscillate about 1:7 chance in the general election.  Similar to Hillary Clinton, Nate Silver repeatedly accepts no blame, but instead daydreams about forgiveness, short memories, or both.  Superciliously claiming Donald Trump's result was merely a 5σ probability (1:1200 year event).  But conveniently of course, no nation on earth has been around that long!

The only moral of the story is it is better to be smart, then it is pretend to be smart.  See the conceited 99% on the left of the chart above?  That's from my follower, Princeton lecturer Sam Wang, who later went on with another, similarly bizarre decision: ingesting a parasite on CNN (who per chart above was no better). 

Work hard to instead 99% of the time, embrace and respect one another's point of view.  Be gratified to be alive in this great nation, and boundless things will still happen.  Life will be too short to moan and live it like this one-shot wonder:




Subscribe by email:


  

No comments:

Post a Comment