3 thoughts on “If you’re at the Joint Statistical Meetings

  1. Is this book out yet? Amazon supposedly has it for overnight delivery, but the web page says it's to be available on September 10.

  2. I saw your talk. When I first saw the median voter theory I thought, "but what if location does not completely detirmine whether someone votes for the candidates. The alternative idea is that it detirmines the expected probability that someone votes for the candidate. Depending on the shape of that curve, what happens in that case is a little bit of difference between candidates makes only a little bit of difference in how people vote. Which sounds more realistic then the fully detirmined vote notion of the median voter theory.

    Then optimum location would be biased towards the other candidate, but still owing quite a bit towards the center.

    In the end you tittled this 'error' a bit later. At first I thought, 'error?'. But then I remembered this is statistics and error is everything we don't model =)

    But still if you discard error and reconstruct it as a hierachical model I think it might go further faster. And also lends itself to the question can we estimate:
    the expected probability that you will vote for a candidate with 'political distance' as a covariate

    All and all it was interesting. You are a pretty good speaker. But man your talk was nothing like the first talk. Quite a roller coast ride we had going in that room.

    On a side note, the first speaker managed to reinvent 'obviously' as the phrase 'from college level physics'. I thought that was funny.

  3. John: The official release date is 10 Sept but the books already exist and are at Amazon and our local bookstore. The publisher sets the official release date later because they never know exactly when the book will get in the stores, and they don't want reviews coming out before the book is available.

    Jeremiah: Just "pretty good," huh?
    Also, regarding your comments on the median voter theorem, yes, I was loosely referring to random utility models as error models. Some of the sorts of models you discuss have been considered in the poli sci literature, I think; the key difficulty is estimating the positions of the candidates and voters based on whatever survey data happen to be available.

Comments are closed.