Tuesday, January 15, 2008

Keith numbers

Election fever is once again starting to boil our collective brain in its natural hardshell container. For some of us, that means a chance to watch the media play with poll numbers. Usually it's a lesson in how not to think about math — see the New Hampshire Democratic results versus the press coverage of polls in the week before, for a recent example. Happily, we now have pollster.com, where we have good social scientists (including UW's Political Science prof Charles Franklin) giving lots of poll data with good commentary on what it all does and does not mean.

In the mainstream media, it looks like Keith Olbermann has taken up the battle to raise the standard. He's proposed the "Keith number" as a guide to reading poll numbers:
What, you ask is the ‘Keith number‘? This is the margin of error plus the percentage of undecided …. I thought of it so, I named it after myself. You think of a better caveat for polls from now on and we‘ll name it after you.
Nice basic point, of course, for reckoning what is a real lead versus the illusion of a lead in a poll. And even pollster.com doesn't always make it easy to see the number of undecided.

But what gets me is that there's an established mathematical use of "Keith number", as detailed here and named after Mike Keith. (It's sort of like a Fibonacci sequence.) Is this a clever inside joke by Olbermann or an accident?

No comments: