Skip to main content

View Diary: A supposedly objective take on Nate Silver; worth the read (56 comments)

Comment Preferences

  •  What's the fuss about? (14+ / 0-)

    538's projections are just slightly modified poll aggregates (with the modification coming in the form of weighting polls based on their past accuracy and current outlier status, among other common sense things). The widespread validation of some "controversy" over the projections eludes me. It's not some mathematical parlor trick or wizardry, it's polling, which in virtually every other circumstance is treated as the "real" state of the race over and above spin.

    it fitfully blows, half conceals, half discloses

    by Addison on Sun Nov 04, 2012 at 09:39:03 AM PST

    •  but he's got ad hominem arguments (0+ / 0-)

      and a poor understanding of what predictions made by statistical models mean!

      •  Well... (0+ / 0-)

        ...he admits to having a long-running grudge against Silver, and so a lot of this article is just the continuation of his vendetta. But I suppose the failure of Silver to get the UK election even close to right says something about something. I just don't know that it's at all applicable to the US race, which is fairly straightforward in terms of projection.

        it fitfully blows, half conceals, half discloses

        by Addison on Sun Nov 04, 2012 at 10:05:50 AM PST

        [ Parent ]

        •  the models are limited by the amount of data (1+ / 0-)
          Recommended by:
          kyril

          available. If you are trying to model hundreds of individual races with a small amount of polling data for each race, the model will probably be fairly useless.

          But if you're mostly focusing on a small group of states that are being polled by multiple firms every day, you have much firmer ground to build on.

          I don't think Silver would disagree with this assessment.

        •  Yes, and (1+ / 0-)
          Recommended by:
          Stroszek

          he makes the false conclusion that trial and error means just guesswork.  Wrong.  It's the scientific method.  Silver grants some polls more weight than others because they have had more accurate results in prior elections, and he always explains that the nuances in his methodology are intended to generate the most accurate results this time around.

          Trial and error is a blunt criticism wielded by these anti-science guys.  How the hell do they think entrepreneurs and inventors work?

          Nobody puts Baby in a binder!

          by deminva on Sun Nov 04, 2012 at 10:16:50 AM PST

          [ Parent ]

    •  Actually, that IS Cosh's argument, (2+ / 0-)
      Recommended by:
      Garrett, kyril

      except that the model Silver uses is both byzantine and proprietary (so no one can analyze it), yet it comes up with numbers not substantially different from very simple poll aggregators, and with the same flaws.   Cosh isn't the first person to argue this (Sam Wang has similar complaints), and it's at least a legitimate angle of criticism - unlike the "We don't like what he's producing" arguments that Silver's been subjected to over the last few months.

      Saint, n. A dead sinner revised and edited. - Ambrose Bierce

      by pico on Sun Nov 04, 2012 at 10:11:14 AM PST

      [ Parent ]

      •  That's ONE thread of his argument... (0+ / 0-)

        ...if you read the linked article (and his past articles) there are a variety of issues Cosh has with Silver, occasionally even indulging it a little bit of Kettle Logic, in my opinion.

        it fitfully blows, half conceals, half discloses

        by Addison on Sun Nov 04, 2012 at 11:16:56 AM PST

        [ Parent ]

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site