Suppose $\bar X$ (respectively $ \bar Y$) is an $ R^m$ (resp. $R^n$) valued random vector with density f(.) and g(.). Let $\phi: R^n \rightarrow R^m$ be $C_1$. If X,Y are independent, then show that $(\bar X + \phi(\bar Y ) | \bar Y = \bar a) \rightarrow^d \bar X + \phi(\bar a)$.
I thought of using convolutions, but how would that even work in n (m?) dimensions? Would the density of $\phi(Y)$ be $g(\phi ^{-1} (X)). J(\phi^{-1})$? I can't seem to wrap my head around it.