- same as an expected value, but weighted by the conditional mass or density instead of the marginal
E(X∣Y=y)=x∑xpX∣Y(x∣y),E(X∣Y=y)=∫−∞∞xfX∣Y(x∣y)dx
- $f_{X \mid Y}(x \mid y) = \frac{f_{X,Y}(x,y)}{f_Y(y)}$, the continuous version of [[Conditional Probability]]
- properties
- linearity: E(aX+bZ∣Y)=aE(X∣Y)+bE(Z∣Y)
- tower law: E(E(X∣Y))=E(X)
- taking out what’s known: for g a function of Y, E(g(Y)X∣Y)=g(Y)E(X∣Y)
- if X≥0 then E(X∣Y)≥0
- example: (X,Y) uniform on the triangle with corners (0,0), (2,0), (0,2), find E(Y∣X=21)
- think of it as a wedge of cheese, a flat density of 21 ( the triangle has area 2 )
- fX(x)=∫02−x21dy=22−x, so fY∣X(y∣x)=2−x1, uniform on [0,2−x]
E(Y∣X=21)=∫03/2y⋅32dy=322y203/2=43
- example ( exercise 40.11 ): two independent appointments, each exponential with mean 20 min ( λ=0.05 ); they total 38 min, how long was the first?
- Y=X1+X2, joint density fX1,X2=λ2e−λ(x1+x2), sum density ( Erlang-2 ) fY(y)=λ2ye−λy
- conditional fX1∣Y(x1∣y)=λ2ye−λyλ2e−λy=y1, uniform on 0<x1<y
- E(X1∣Y=y)=2y, so 19 min, which symmetry already says