Skip to main content

View Diary: Retired NSA Analyst Proves GOP Is Stealing Elections (203 comments)

Comment Preferences

  •  So you're saying the method used is bs? (0+ / 0-)

    I really want to hear from people who do stats for a living/avocation to confirm this (not saying you don't, just the more who know of this, the better).  But why would Romney's numbers climb, Paul's numbers drop, and the other two candidates' numbers stay steady?

    In capitalist America, bank robs you!

    by madhaus on Fri Oct 26, 2012 at 07:51:24 PM PDT

    [ Parent ]

    •  that's not exactly what I'm saying (1+ / 0-)
      Recommended by:
      Bob Love

      The method used here is one legitimate way of visualizing the relationship between votes cast per precinct and vote shares. Another way would be to fit a smoother to the scatterplot -- although arguably that would give too much visual emphasis to "small" precincts relative to larger ones.

      (I used scare quotes because votes cast in the Republican primary isn't exactly a size metric. I've been in some 98% Democratic precincts.)

      The big honking fallacy here is that an upward (or downward) slope is prima facie evidence of fraud.  That would only make sense if voters were randomly distributed -- or, at least, the number of Republican voters per precinct were.

      What explains this relationship? One possibility is that Republicans in precincts with many Republicans are more likely to support "establishment" candidates. But how do we measure people's propensity to vote for "establishment" Republicans, at the precinct level? We'd probably do it by looking at past voting behavior. And when we did that, the Paulites would say, "See? the establishment has been screwing us for years!!!1!" Even though the correlation was originally asserted to be an "anomaly," the argument can morph ad libitum to fit the empirical data. Trying to prove that the correlation has innocent causes is like trying to convince someone that Obama isn't a socialist.

      (It used to be that people would say that exit polls were uncannily accurate predictors until 2004 -- or, a few would say, until 2000. That was demonstrably false. Steve Freeman, in his book, has a list of all the known 'exceptions' to the 'rule' that reads like an entry in a satirical No True Scotsman contest. People who haven't gotten that far still say that exit polls were uncannily accurate predictors until 2000 or 2004. It's a zombie undead non-fact. That aside, Freeman managed to accommodate the facts, to some extent, without altering his opinion in any observable way.)

      So, I step back and say: (1) Choquette and Johnson's assumption makes no sense. (2) Their interpretation makes no sense: It is ridiculous to propose that Romney stole votes in a very large fraction of precincts, in 49 of 50 states using a wide variety of election technologies (including hand counts), favoring the precincts with more votes in order to reduce the chance of detection. How dumb does an argument have to be before people realize that it doesn't have to be rebutted mathematically? (That isn't directed at you; that's a question that has troubled me for years.)

      Election protection: there's an app for that! -- and a toll-free hotline: 866-OUR-VOTE
      Better Know Your Voting System with the Verifier!

      by HudsonValleyMark on Sat Oct 27, 2012 at 05:34:07 AM PDT

      [ Parent ]

      •  It's not just a visualization ... (0+ / 0-)

        Charts do tell a lot, but there is also a strong statistical study that backs up the work.

        We have factored out urban density in 11 states that provided precinct-level data and the anomaly remained. Look at the statistical analysis or show it to someone who can appreciate such work.

        We go to great lengths to prove the randomness of the data, using 5 different tests that indicate randomness. All the tests pass, yet the anomaly remains.

        This second paper backs up the evidence in the first one:
        http://www.themoneyparty.org/...

        •  I've been looking at this since August (0+ / 0-)

          There's no there there. "Tests that indicate randomness"?! It's voting data, for pete's sake. The null hypothesis makes no sense, there's no good reason to expect to have measures of the causal variables, and the rationale for why the relationship evinces fraud is incoherent. I'm sorry, but you will have to do better, and I don't think you can.

          Election protection: there's an app for that! -- and a toll-free hotline: 866-OUR-VOTE
          Better Know Your Voting System with the Verifier!

          by HudsonValleyMark on Sun Nov 04, 2012 at 04:26:18 PM PST

          [ Parent ]

        •  "tests that indicate randomness" (0+ / 0-)

          Define "randomness" in this context.

          Thanks :)

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site