Skip to main content

View Diary: Cell Phone Users For Obama, Land Liners For Romney (25 comments)

Comment Preferences

  •  spurious means it appears to be valid but isn't: (1+ / 0-)
    Recommended by:

    For example the correlation between the foraging activities of squirrels in Brazil and the presence of students on the Cal Berkeley campus.

    The value of correlations is that they can be used to make accurate predictions.  Spurious correlations lead to less-accurate predictions.  

    If you want to predict the best times of year to advertise to students at Cal Berkeley, you don't want to choose your ad buys based on the squirrel activity in Brazil.  What you want to use is the academic calendar at Cal Berkeley.  

    "Minus two votes for the Democrat" equals "plus one vote for the Republican." Arithmetic doesn't care about your feelings.

    by G2geek on Fri Sep 07, 2012 at 09:30:08 PM PDT

    [ Parent ]

    •  I know what spurious means (1+ / 0-)
      Recommended by:

      I didn't see a claim of causal relationship in the first poster's theory either. Just correlation.

      Hence my question.

      "Well, the problem here is that you're out of candy. You're gonna need more candy." Rachel Maddow on the Big Bailout

      by cishart on Sat Sep 08, 2012 at 12:02:25 AM PDT

      [ Parent ]

      •  right, but my point is that the correlation.... (2+ / 0-)
        Recommended by:
        Sophie Amrain, cachola

        .... is spurious, like Cal Berkeley students and Brazilian squirrels.

        For example there is a plurality of geeks and others who don't use cellphones because we don't want to be tracked, monitored, and spied on.  In some parts of the US, those numbers tilt toward the left, therefore progressive, and in other parts of the US they tilt toward libertarian, the latter usually lumped in with conservative.

        There is also a plurality of people who don't use cellphones because they prefer simpler technology, or they're skittish about microwaving their brains.  These can tilt either toward the left or the right depending on which part of the US one is looking at.

        A pollster who is attempting to make a generalization about political orientations based on type of telephone service, is making inferences based on correlations that have enough exceptions as to be effectively spurious.  

        Those pollsters are attaching a meaning where it doesn't really fit.  It's a spurious correlation because it's weak compared to other correlations that deal with the same variables and are stronger.

        Now of course we want our pollsters to poll "cellphone only" users and households, just as we want them to poll "landline only" subscribers, and "both landline and cellphone" subscribers.  And for that matter we want them to poll households that don't have any kind of phones!  We don't want our polling data artificially restricted on the basis of the way the poll is conducted.

        But neither do we want pollsters making predictions based on correlations that are weak enough as to be misleading.  If that's what they want to do, they should go into the astrology business.  Really.

        "Minus two votes for the Democrat" equals "plus one vote for the Republican." Arithmetic doesn't care about your feelings.

        by G2geek on Sat Sep 08, 2012 at 12:33:55 AM PDT

        [ Parent ]

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site