Skip to main content

View Diary: Hey Zimmerman, Most Thieves Are White With Video And Graphs (52 comments)

Comment Preferences

  •  Not racist, just not understanding the gambler's (0+ / 0-)

    fallacy.

    Nor the idea that you can't generalize from one person to another. Especially not based on one case each.

    Ou sont les neigedens d'antan?

    by SouthernLiberalinMD on Fri Jul 12, 2013 at 11:47:30 AM PDT

    [ Parent ]

    •  Lol (0+ / 0-)

      What about the Gambler's fallacy do I "not understand"?

      Nor the idea that you can't generalize from one person to another. Especially not based on one case each.
      I'll take "missing the point" for $200, Alex.

      (-5.50,-6.67): Left Libertarian
      Leadership doesn't mean taking a straw poll and then just throwing up your hands. -Jyrinx

      by Sparhawk on Fri Jul 12, 2013 at 12:01:19 PM PDT

      [ Parent ]

      •  past rolls of dice don't determine future rolls (0+ / 0-)

        and one pink person's action doesn't determine the actions of the next pink person you meet.

        Ou sont les neigedens d'antan?

        by SouthernLiberalinMD on Fri Jul 12, 2013 at 12:07:13 PM PDT

        [ Parent ]

        •  The Fallacy of the Gamblers' Fallacy (0+ / 0-)

          This comment sequence is a good demo why devolving into the world of marbles can be foolish when we were working with a perfectly good example from real life (see other comments on this diary).

          In the data from Sparhawk you have available, the risk of being of a thief is associated with being blue.  The more data you have, like an entire statewide database, the better your predictions.  If you re-sample another person, you are selecting from within the sampling frame that has already been used to determine the risk of thievery.  So the average risk for a blue is higher than a pink.  

          The true probability for any single blue or pink is either 0 or 1.0--meaning, there are determining and unknown (or unavailable to the observer) factors other than blueness or pinkness (which are almost certainly not causal, but merely are just correlated).  But if all you have to go on is the color, then it is mathematically correct to assume that  any individual blue is, on average, more likely to be a thief.

          A lot of these issues are really about making informal Bayesian predictions, and humans aren't always very good at that.

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site