November 25, 2008

Quick Back-Pat Update

SourceObama
McCainDem Error
Rep Error
Total Error
Election 200852.845.9---------
RCP52.144.50.71.42.1
53852.346.20.50.30.8
Chris53.145.90.30.00.3
Posted by at
Labels: , , ,

3 comments:

  1. BartNovember 25, 2008 5:07 PM

    You, sir, have a legitimate shot of writing for 538 before the next election cycle really begins. Seriously, you should send Nate Silver an email that includes a description of your model, a link to your posts on our blog, and an attached photograph of you taunting him for having a worse model than yours. I think you would be hired immediately.

    ReplyDelete
  2. ChrisNovember 25, 2008 5:19 PM

    I'm leaving work so I'll rebut this point in more detail later. But I'll just put it this way: leading up to the election I'd say it's fair the consensus was that Obama was ahead by about 7 pts. In 2004, 99% of voters voted either for Kerry or Bush.

    Pick two numbers that have a sum of 99 and a difference of 7. Answer? 53-46.
    (53.0-52.8)+(46-45.9)=0.3
    See? Anyone could have done just as well!

    ReplyDelete
  3. ChrisFebruary 03, 2009 4:52 PM

    (I'm finally getting around to analyzing my data and I found this post when I needed to know how to do some HTML)

    The other point being that all I did is I took his model, removed 90% of the rigor, and got lucky with the rest.

    The strongest part of my model is giving each poll a weight based on its half-life and the pollster's previous success. I used his half-life formula and just changed the half-life to whatever looked good, and directly stole his pollster weights.

    ... finally, the stuff he's now doing re: the Senate is way more work than I'm willing to put in. What I don't understand is why he's doing this for free, or at least a small fraction of what he could make as a consultant.

    ReplyDelete

Note: Only a member of this blog may post a comment.

Subscribe to: Post Comments (Atom)