Skip to main content

View Diary: Nate Silver Calls B.S. on U. of Colorado Election Prediction Model UPDATED x1 (139 comments)

Comment Preferences

  •  Deeper than that, even models w predictive value (2+ / 0-)
    Recommended by:
    SneakySnu, ericlewis0

    do not necessarily have fidelity to the underlying causative dynamics.

    So even "proven" predictive value is relative, and still quite limited, and can fail next time around, because it doesn't model enough of the nonlinear dynamics.

    And even those dynamic models that do contain causative feedback networks are limited by emergent phenomena that were not contained within the scope of the model, but which intrude into the corresponding scope within the real world. Think Butterfly Effect, and then generalize. Emergent phenomena can bubble-up from levels of scale beneath the resolution threshold of the model. They can come in from outside the spatial scope of the model ("out of left field"). They can come from outside the temporal scope of the model.

    No model is perfect. Period.

    But some models are much worse than others.

    Thanks Nate and Eric for pointing out this one as a blatantly-bad non-starter.

    #3: ensure network neutrality; #2: ensure electoral integrity; #1: ensure ecosystemic sustainability.

    by ivote2004 on Fri Aug 24, 2012 at 12:40:24 AM PDT

    [ Parent ]

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site