Skip to main content

View Diary: Who Will Win The Toss Up States? (66 comments)

Comment Preferences

  •  I didn't think there were any toss up states (2+ / 0-)

    McPalin has four and Barack has the rest.

    Sarah Palin is just another Karl Rove in lipstick and stilettos.

    by uc booker on Tue Oct 14, 2008 at 08:30:07 AM PDT

    •  There are 4, I think (2+ / 0-)

      Sam Wang has the cleanest model of this, and last I checked only 4 states have an Obama win probability between .05 and .95.

      That's why this is already over.

      Manny goes to Hollywood :(

      by theran on Tue Oct 14, 2008 at 08:55:46 AM PDT

      [ Parent ]

      •  A lot depends on the operational (2+ / 0-)
        Recommended by:
        theran, crazyshirley2100

        definition of "toss up" that one factor changes your model a lot. I have been using Pollster for the whole cycle so I stick with them and their fairly conservative model.

        Every word Republicans say is a lie, including "And" and "The".

        by Something the Dog Said on Tue Oct 14, 2008 at 08:57:16 AM PDT

        [ Parent ]

        •  I put less stock in these loess trend lines (2+ / 0-)

          than others do, since they tend to blur the current state of things, and it's not clear how to weight the past.  Basically, taking the median of recent polls to estimate win probability seems to be a good way to deal with the inherent variation in them.

          The trend lines are trying to capture more information than you need to decide who would win today, and they don't predict future ``game changers,'' so I've decided that I am partial to just looking at the snapshots.

          That said, I can see the argument for the other side.  (I can't really buy into a lot of the tightening assumptions and dependency models in 538, since they don't really have enough training data.  Essentially it's converging to the snapshot approach anyway.

          The 538 blog is awesome, though, as was the poll-free predictions from the primary.)

          Manny goes to Hollywood :(

          by theran on Tue Oct 14, 2008 at 09:04:19 AM PDT

          [ Parent ]

          •  Yeah, (2+ / 0-)
            Recommended by:
            theran, crazyshirley2100

            I come down on the side of trends since I use them a lot in my work, but there is still a lot of voodoo in the works when you use them for future predictions.

            Every word Republicans say is a lie, including "And" and "The".

            by Something the Dog Said on Tue Oct 14, 2008 at 09:06:22 AM PDT

            [ Parent ]

            •  Non-parametric estimation is great (2+ / 0-)

              if you assume that the underlying system is stable, and that you'll get to see it long enough to converge to something interesting.  (Really---I think non-parametric statistics is one of those beautiful things everybody should be exposed to in undergrad or early grad school.)

              I just don't really think there's as much structure to be mined here, either in the trends or in the dependencies among the states, on the parametric side.

              Of course, this is just a heuristic intuition.

              Manny goes to Hollywood :(

              by theran on Tue Oct 14, 2008 at 09:11:57 AM PDT

              [ Parent ]

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site