Law of iterated expectations

Поделиться
HTML-код
  • Опубликовано: 25 янв 2025

Комментарии • 12

  • @thelateknights
    @thelateknights 4 года назад +31

    This wasn't made explicit in the video, but at 0:44 , those values are derives as follows:
    X = 0 E(Y|X=0) = .1 (this is the interception of y=1 and x=0) / (.4 + .1) (this is the marginal probability of x=0) = .2
    X = 1 E(Y|X=1) = .4 (this is the interception of y=1 and x=1) / (.1 + .4) (this is the marginal probability of x=1) = .8

  • @algorithmo134
    @algorithmo134 2 года назад +2

    At 3:26 how do you get pdf of z

  • @jayjayf9699
    @jayjayf9699 5 лет назад +4

    How did u get expected Value of y given x =0 equal to 0.2?

    • @revysingh
      @revysingh 5 лет назад +2

      That is exactly the same question I had in my head. If x is 0 y is 0 (with a probability of .4) or 1 (with a probability of .1). How does this end up being an expected value of 0.2?

    • @jayjayf9699
      @jayjayf9699 5 лет назад +2

      Revy Singh it is the correct answer from the definition of conditional expectation we have the suming over the y values for example for given x=0 we get 0*(0.4/0.5) + 1*(0.1/0.5) =0.2 ......I got the 0.5 from adding 0.4+0.1 as this give the probability of x=0

  • @Agustinoism
    @Agustinoism 11 месяцев назад

    guys there is a sligt error
    E(Y|x=x)=y*f(y|x)dy

  • @timurgareev6356
    @timurgareev6356 3 года назад

    Thank you, could you clarify how to get f(z)?

    • @yaweli2968
      @yaweli2968 3 года назад

      It’s fz(Z)={0.5 ,if z=0.2 or z=0.8. Remember Z is always E(Y|X) and those two (0.2, 0.8) are the only possible values of Z. So Z has a 50% chance of being one of the two at a time.

    • @timurgareev6356
      @timurgareev6356 3 года назад +2

      @@yaweli2968 I still don't understand where did he get 1/(18 (2 z - 1)^3) ? (timing i.e. 4:23)

  • @harshalverma5757
    @harshalverma5757 4 года назад +1

    as said at 8:15 how fy(y) = y+1/2

    • @andrewfox2480
      @andrewfox2480 4 года назад +1

      You get that result when you integrate the joint density function f(x,y) with respect to x, while setting the limits of the integration equal to the lower bound and upper bound limits of x (0 and 1). By doing this, you integrate the random variable X out of the joint density function, and you are left with the marginal density function (or the probability density function) of just the random variable Y.

  • @yt-1161
    @yt-1161 2 года назад +1

    if z=g(x) and g is monotone decreasing then F_Z(z)=P(Z = g^-1(z))= 1 - F_X(g^-1(z))
    so f_Z(z)= -d/dz (F_X(g^-1(z))) = - f_X(g^-1(x)) d/dz g^-1(z) (main formula)
    here if Z = (3X+2)/(6X+3) =: g(X) then X = (-3+Z)/(6-3/Z) = g^-1(Z)
    in the last video he computed marginal of X from the joint x+y, it was f_X(x) = x+1/2
    so insert it in the main formula
    -((-3+z)/(6-3/z) + 1/2) * ((-3+z)/(6-3/z))' = 1/(18 (-1 + 2 z)^3)
    check it here
    www.wolframalpha.com/input?i=-%28%28-3%2B2%2Fz%29%2F%286-3%2Fz%29%2B1%2F2%29*%28-1%2F%283+%281+-+2+z%29%5E2%29%29