lukestein’s avatarlukestein’s Twitter Archive—№ 7,272

  1. …in reply to @paulgp
    @paulgp (Assuming I understand what you’re doing) if it were a panel regression with state FE, yes (since the log and FE would pick up the population normalization). In a cross-sectional regression, the log helps but doesn’t solve at all.
    On twitter.com Twitter logo lukestein’s avatar ♻️ 2 Retweets Mood +4 🙂
    1. …in reply to @lukestein
      @paulgp Imagine instead of your good fancy model, my model was to assume that each state’s claims were 2% of the population if name starts with a consonant, 4% of population if it starts with a vowel. I’m going to get a big R2 in cross-sectional regressions with or without log, right?
      Permalink On twitter.com Twitter logo lukestein’s avatar ♻️ 2 Retweets ❤️ 1 Favorite Mood +4 🙂