Skip navigation.
The Texas Blue
Advancing Progressive Ideas

Franken By 27?

Who does regression analysis on a recount, for goodness' sake?

I suppose going to work is much less exciting for pollsters these days than it was a few short weeks ago, but Nate Silver isn't ready for vacay time, apparently. He's been looking closely at the day-to-day recount numbers in the Minnesota and has noted that on days where Franken has netted fewer votes compared to Election Day than Coleman, there were also more challenges to the vote. Silver used this fact to come up with a fairly complex regression that projects that after all contested ballots are judged, Franken will end up winning by 27 votes.

Silver points out that the error on these calculations could be high, though he thinks the correlation he's identified is strong. Obviously, I'm hoping he's right, but the big winner here is Nate Silver -- if he's wrong, the story dies and is buried; if he's right, he will instantaneously be the most famous pollster in America.

Me! Pick Me!

*raises hand*


Actually, the sheer drop in cost of computing power makes this kind of thing a lot more accessible to a lot of people. I'm really looking forward to seeing people develop new applications of methods that were previously only available to commercial data analysts. -dx

Dheeraj Chand is a freelance journalist, blogger, data analyst and writer in Washington, DC. He is from Houston, TX, originally.

In Addition

In addition to the relative affordability of computing power, there's also more tools available as well. Things like R and Ggobi and GRASS GIS place a tremendous amount of affordable statistical power in individuals' hands.

The Model

The only remaining hurdle to cross, then, with putting that sort of "statistical power in individuals' hands" is that, while fitting a model is made that much easier, judging the quality of a model is still as difficult as it has always been. Any grad student can do a regression. :-) Whether that regression has any real significance is dependent on a person's ability to put a microcosm of the world into mathematical terms -- and I'm not sure that'll *ever* get much easier.

Syndicate content