Let $(\mathcal{X},\mathcal{A})$ and $(\mathcal{Y},\mathcal{B})$ are Borel spaces.
Let $\kappa_i: \mathcal{B} \times \mathcal{X} \to [0,1], ~i=1,2$ be Marcov kernels from $(\mathcal{X},\mathcal{A})$ to $(\mathcal{Y},\mathcal{B})$, i.e., for each $B \in \mathcal{B}$, $x \mapsto \kappa_i(B, x)$ is a $\mathcal{A}$-measurable function, and for each $x \in \mathcal{X}$, $B \mapsto \kappa_i(B, x)$ is a probability measure on $(\mathcal{Y}, \mathcal{B})$.
Suppose that for any $x \in \mathcal{X}$, $\kappa_1(\cdot, x) \ll \kappa_2(\cdot, x)$ and let $f(x,y)$ denote the Radon-Nikodym derivative $\frac{d\kappa_1(\cdot, x)}{d\kappa_2(\cdot, x)}(y)$.
My question: is it true that $(x,y) \mapsto f(x,y)$ is $\mathcal{A} \otimes \mathcal{B}$-measurable?
I am not familiar with the Marcov kernel, but this question is motivated by that I want to resolve my previous question.
Any comments are welcome. Thank you very much.
The condition that $\mathcal{B}$ is separable is critical.
– jet457 Mar 23 '25 at 04:18