Rob Kall OpEdNews.com
"Wow, you're accurate." That's what Kathy Lash, deputy press secretary for Howard Dean said in her call to me a few minutes ago. "How'd you deal with all that data so fast?" she reported that her candidate, Howard Dean asked.
I knew what she was talking about-- one of the stats I'd used when I analyzed the raw data a bit differently predicted that Dean would get exactly 43% of the vote.He actually received 43.8. I'd eyeballed the data and told the Dean crew yesterday that he'd won, but I didn't think he'd pulled the 50% it would take to get money from Moveon.org. Still, the word was that Dean's campaign manager was very happy last night.
That's because I'd called the Dean campaign to let them know the "internet exit pollTM" results I'd calculated showed Dean to be the winner-- and I was telling them the day before the official results were being released, a day before the official "exit poll" done by calling 400 voters was released. And.. I'd used data from hundreds of thousands of web users. .
So... how did I do?
What I got right: I correctly predicted the winner, the first three, the first four, the loser, the fact that Kucinich was in second place for part of the time, the fact that Carol Mosely-Braun had made a good showing. Actually, in the post election data analysis I found that the bounce went flat for all of the candidates except Mosely-Braun, whose stats actually got better.
What I got wrong: I had Kerry ahead of Kucinich. The one stat that put Kucinich in second on the first day of the primary was the same stat that predicted Dean's slot within less than one percent. The rest of the candidates-- they were all so close in the final tally, except for Sharpton and Lieberman. Actually, our day after stats clearly showed that Lieberman had dropped drastically. One factor that contributes to our prediction errors was explained by our discussion with alexa corporation, the company that we used for stats. They explained that the lower the number ranking, the more reliable the numbers are, the more accurate they are. There's more "looseness" in the numbers when you go higher rankings and that's what we had for the people in the 5-8 rankings. Of course the biggest question was "Who won?" and we nailed that one.
This is the first time we actually have real numbers to work with, to compare to our stats. That will help us to be more accurate the next time around. We're hoping to be able to work with some of the data moveon.org has. This might enable us to take this nascent field to the next level.
Rob Kall rob@opednews.com is publisher of progressive news and opinion website www.opednews.com and organizer of cutting edge meetings that bring together world leaders, such as the Winter Brain Meeting and the StoryCon Summit Meeting on the Art, Science and Application of Story. He has also been co-developer of several software programs, and inventor of one of the most widely selling biofeedback devices. This article is copyright by Rob Kall, but permission is granted for reprint in print, email, blog, or web media so long as this credit is attached
