OT - Happy Election Day!

[Vadered]

I love the projections. Less than one percent reporting in NJ, Romney has a 14% lead in the early returns, and FOX is calling it for Obama.

Statistics are weird.

 

[Dacks]

I love the projections. Less than one percent reporting in NJ, Romney has a 14% lead in the early returns, and FOX is calling it for Obama.

Statistics are weird.

Yep. Even after working at Statscan for a few years, the inferences you can make from very small sample sizes is remarkable.

They're only legitimate if the math is solid and the sample is truly random and/or representative. (And even then they're only accurate x% of the time.) So, there's always a risk that that one percent is biased, i.e. maybe it comes from low-density areas (which might finish polling early) or from early voting. If those voters trend a certain way, than the early votes are definitely biased. But otherwise, you don't need much to make a good prediction.

Here would be a good experiment for the pollers. Make the statistican predict results (Obama vs. Romney) the moment the polling meets their 95% confidence level threshold. At the end of the day, see how well those predictions panned out. If more than 95% were correct, they've done things properly. If less, they're making a mistake somewhere. I can guarantee they're way under.

Math lecture over.

 

[Vadered]

Yep. Even after working at Statscan for a few years, the inferences you can make from very small sample sizes is remarkable.

They're only legitimate if the math is solid and the sample is truly random and/or representative. (And even then they're only accurate x% of the time.) So, there's always a risk that that one percent is biased, i.e. maybe it comes from low-density areas (which might finish polling early) or from early voting. If those voters trend a certain way, than the early votes are definitely biased. But otherwise, you don't need much to make a good prediction.

Here would be a good experiment for the pollers. Make the statistican predict results (Obama vs. Romney) the moment the polling meets their 95% confidence level threshold. At the end of the day, see how well those predictions panned out. If more than 95% were correct, they've done things properly. If less, they're making a mistake somewhere. I can guarantee they're way under.

Math lecture over.

The problem I have with it is that I have a hard time believing that the one percent is really random. I imagine it comes from one or two counties, which might be predominantly one side or the other, and even if they aren't, there is no statistical metric on earth which says this guy has 43% of the vote therefore he will beat the guy with 57% of the vote. Gotta be A) a lot of behind the scenes assumptions involved and/or B) some amount of making shit up.

 

[Dacks]

The problem I have with it is that I have a hard time believing that the one percent is really random. I imagine it comes from one or two counties, which might be predominantly one side or the other, and even if they aren't, there is no statistical metric on earth which says this guy has 43% of the vote therefore he will beat the guy with 57% of the vote. Gotta be A) a lot of behind the scenes assumptions involved and/or B) some amount of making shit up.

Yeah, it sounds like in that example they are using a small, unrepresentative sample to make inferences about the population. In that case, the inferences are less statistically based, and probably a historical inference. (For example, if the democrat consistently wins 10% more of the state-wide vote than he does in that county.) That becomes quite a big assumption. But everyone wants to be the first to call it, and hope that everyone forgets their mistakes by the end of the night.

 

[Vadered]

Hey, Florida is within 1000 votes again.

Forget Dallas, fuck Florida.

 

[dash]

Hey, Florida is within 1000 votes again.

Forget Dallas, fuck Florida.

That guy named Chad gumming up the works again...

 

[Winged_Wheel88]

At the moment, this feels like an Obama night tonight.

 

[Cricket]

Is it over yet?

 

[Winged_Wheel88]

SHE'S NOT DEAD

 

[puckhead]

holy shit, Zombie Cricket again!
welcome.

 

[Winged_Wheel88]

, Cricket

 

[IPostedWhat]

No matter who wins, the American people are still the losers.

 

[Cricket]

Hey y'all.

 

[sabresfaninthesouth]

I love the projections. Less than one percent reporting in NJ, Romney has a 14% lead in the early returns, and FOX is calling it for Obama.

Statistics are weird.

Some states are just so easy to predict. If the first precinct out of California is a heavily Republican precinct, the early count would show Romney ahead, but that's not going to stop the talking heads from calling it for Obama.

 

[Jimmy_the_Tongue]

Romney is toast... but Nixon lost against JFK in the early '60's and went on to be president....and that turned out ok!!!!

 

[Dacks]

Well, they haven't called Florida or Ohio yet, but CNN's county breakdown seems to imply the late votes will go to Obama. Short of some major recounts or spoiled ballots, I think it's a second term for 'Bronco Bama.

 

[Winged_Wheel88]

Well, they haven't called Florida or Ohio yet, but CNN's county breakdown seems to imply the late votes will go to Obama. Short of some major recounts or spoiled ballots, I think it's a second term for 'Bronco Bama.

Jesus Christ, I had thought I had heard every pun on his name in the last 4 years and you hit me with a new one!

 

[IPostedWhat]

Just think of all the people that could of been helped out with the money that was wasted by both parties on attack ads.

 

[Winged_Wheel88]

Just think of all the people that could of been helped out with the money that was wasted by both parties on attack ads.

That thought had already crossed my mind tonight.

 

[Winged_Wheel88]

NC went Romney. The race is still up in the air.