Monday, November 24, 2008

Stats don't lie

Data can be misinterpreted, but it never lies. Quality interpretation usually leads to plenty of "I told you so's" while poor interpretation be it intentionally, ignorantly, or just through bias, either gets ignored, qualified (there was no way to foresee that happening), or posthumously changed.

Faux News usually provides great examples of poor performance. Remember all the calls from them that McCain was "closing the gap" from 7 points behind Obama in the days preceding the election? The final results left Obama 53%-46%, dead on the 7 points predicted by most reputable sources.

On the other side, two of my favorite sources through the election were Electoral-Vote.com and FiveThirtyEight.com, both of which only missed Indiana and the Omaha district of Nebraska (FiveThirtyEight did not make a prediction for Missouri) in their pre-election predictions.

Maps from Electoral-Vote.com and FiveThirtyEight.com (respectively):


And the final map (from CNN.com):



Regardless of what the naysayers conclude about the illegitimacy of polls, a quality interpretation of the data will yield a correct result, as demonstrated by E-V.com and 538.com.

2 comments:

Drew said...

The thing is, there are two kinds of people in the world: Those who observe the world, test it, and draw logically-based conclusions from the observable data; and those who will always try to bend the data to conform to their pre-existing beliefs.

The first kind of thinking we call "science"; the second, "wishful thinking." Science: It Works, Bitches.

shine said...
This comment has been removed by the author.