General Claim: If (X, Y) ⊥ Z, then Y ⊥ Z and X ⊥ Z. True or False?
If false, what about if (X,Y) are jointly Gaussian?
Specific problem I am dealing with: Consider a case of simple linear regression, where $$y_i = B_0 + B_1x_i + \sigma z_i \text{ ~ }_{i.i.d.} N(B_0 + B_1x_i, \sigma^2),i=1,…n$$
There is a result that $$(B_0,B_1)\perp \sum_i (y_i – (B_0 + B_1x_i))^2$$
I want to know if $$B_i\perp \sum_i (y_i – (B_0 + B_1x_i))^2$$
Best Answer
The general statement is correct.
Proof: $$ P(X=x,Z=z) = \int_Y P(X=x,Y=y,Z=z)dy = $$
$$\int_Y P(X=x,Y=y)P(Z=z)dy=P(X=x)P(Z=z) $$