Endogeneity lecture 4: Measurement error and attenuation bias.

Поделиться
HTML-код
  • Опубликовано: 3 ноя 2024

Комментарии • 8

  • @kaungzing1354
    @kaungzing1354 2 года назад

    hi, can i please know is there any circumstances where this type of bias does not matter?

  • @rokassteponavicius4467
    @rokassteponavicius4467 3 года назад

    great explanation!

  • @ryan2009cov
    @ryan2009cov 2 года назад +1

    Why is Var(X1*) = Var(X1) + Var(W) ? Did you show this in the video before proclaiming it at 19:55?

    • @christian.adriano
      @christian.adriano 2 года назад +2

      It comes from the formula for the variance of the sum of two random variables
      Var(X+Y) = Var(X) + 2Cov(X,Y) + Var(Y)
      X1*=X+W
      Var(X1*)= Var(X1+W), applying the formula above,
      Var(X1*)= Var(X1) + 2Cov(X1,W) + Var(W)
      given that Cov(X1,W)=0, because assumption-2 (6:54), i.e., no systematic measurement error, then
      Var(X1*) = Var(X1) + Var(W)

    • @youshengtang3997
      @youshengtang3997 2 года назад

      same question, did you figure it out?

    • @r.barakat8758
      @r.barakat8758 5 месяцев назад

      ​@@youshengtang3997 simply Var(x*)=Var(x+w)=var(x)+var(w)+2Cov(x,w), which equals to var(x)+var(w) since Cov(x,w) equals to zero by assumption. However, I cant prove that Cov(x,w)=var(w) any good?

    • @r.barakat8758
      @r.barakat8758 5 месяцев назад

      ​ @youshengtang3997 simply Var(x*)=Var(x+w)=var(x)+var(w)+2Cov(x,w), which equals to var(x)+var(w) since Cov(x,w) equals to zero by assumption. However, I cant prove that Cov(x,w)=var(w) any good?