Pointwise conditional probability intuition?

26 Views Asked by At

Consider a well-defined joint distribution over $\mathcal{X} \times \mathcal{Y}$. Does

$$P(Y=f(x) \mid X=x) \geq P(Y = g(x) \mid X=x) \ \ \forall x\in\mathcal{X}$$

always imply

$$P(Y=f(X) \mid X=x) \geq P(Y = g(X) \mid X=x)$$ ?

Intuitively, the second inequality should naturally hold but the proof seems non-trivial to me. Any thoughts?

1

There are 1 best solutions below

0
On

The two inequalities are equivalent for any $x$ since

$$P(Y=h(X)\mid X=x)=P(Y=h(x)\mid X=x)$$