GameOver22 said:
I'm not saying he's putting a partisan slant on it. I even admitted there's seems to be a republican bias to the polls....the bias is just much smaller if you look at the later polls (the final week). Generally, when people look at the accuracy of polls, they look at the polls over the final week because it gives a better snapshot of the electorate. Point being, the closer you get to the election, the more accurate the polls tend to be. For instance, an actual peer reviewed article on poll accuaracy from 2008. If you recognize, the earliest poll they consider was released on 10/29....7 days before the election.....not 21 days. The main point is, a lot can happen over three weeks, and a snapshot 3 weeks out is not necessarily an accurate snapshot of election day behavior. |
Pretty much it...
Well that and Nate Silver talked previously negativly about thoughts of polling "conspiracy theories" on the basis that pollsters don't mess with their numbers.
Yet his reason for spreading it out to 21 days ?
"Our method of evaluating pollsters has typically involved looking at all the polls that a firm conducted over the final three weeks of the campaign, rather than its very last poll alone. The reason for this is that some polling firms may engage in “herding” toward the end of the campaign, changing their methods and assumptions such that their results are more in line with those of other polling firms."
Pollsters may be messing with their numbers.
More or less he just picked a longer sample period for to have a wider variance and wider bias to inflate interest and create some more inflamatory numbers.
Talking about errors up to 6% is going to get you more hits and keep peoples attention more then 1.
A more accurate analysis, that would very little extra time since the data is already there would be to run three analysis.
A 21 day, a 7 day, and a final survey analysis.
I mean, if your familar with a statistics program all it takes no time at all to run more analysis off parts of data already entered.
Then you'd even be able to compare them to each other and get a clearer picture of what was going on and figure out what variables likely made the biggest differences.
Heck pretty much anybody who remembers a statisitcs class could run the analysis Silver did with those added parameters... if they wanted to go through the trouble setting up the dataset and of actually doing it.
