• It’s not a bad take - if it shows a consistent bias, it’s still consistent data. It’s translating the bias from a descriptive to a predictive model that’s the hard part. Maybe they found that the swings in correlation were too wide.

    • @Eccitaze
      link
      fedilink
      29 months ago

      IIRC they ejected them because Rasmussen Reports put out a ridiculously flawed article that called the results of the Arizona gubernatorial election into question based on a study whose methodology was so flawed that it could be torn apart by a particularly sharp grade schooler–they took a poll, sponsored by a Republican group, four months after the election, then weighted it against exit polls (not the actual election results), and then used that to claim the Republican won by eight points instead of losing by 1. This prompted the guy in charge of 538 to send them a letter basically saying “are you gonna fix your methodology to reflect something close to reality, orrrrr…” and Rasmussen said “lol no”