Expectation of inner product of independent random variables

hilbert-spacesindependenceorthogonalityprobability theory

Let $H$ be a Hilbert space with inner product $\langle .,.\rangle_H$. Let $X,Y$ be two independent $H$-valued random variables, both of them Bochner-integrable, with $\mathbb{E}\|X\|_H^2 < \infty$, similar for $Y$. Is it true that $\mathbb{E}\langle X,Y\rangle_H = \langle \mathbb{E}[X], \mathbb{E}[Y]\rangle_H$?

This statement is clear to me in the case of $H=\mathbb{R}^n$, but somehow I feel uncomfortable using this in arbitrary Hilbert spaces, so I would be very glad if someone could give me a reference for a proof (or suppy a proof themselves). If that helps, you may assume that $H$ is separable.

Best Answer

Bochner integrable r.v.'s take values inside a separable subspace , so we can certainly assume separability of $H$. Let $(e_n)$ be an orthonormal basis for $H$. We have to show that $E \sum \langle X, e_n \rangle \langle Y, e_n \rangle = \langle EX, EY \rangle$. Note that $EX=\sum E\langle X, e_n \rangle e_n$ $\cdots$ (1). So $\langle EX, EY \rangle=\sum E\langle X, e_n \rangle E\langle Y, e_n \rangle $. The proof is now clear by indepndence provide interchange of sums and expectations i have used can be justified. But the justification is easy using Fubini's Theorem: $E\sum Y_n=\sum EY_n$ if $\sum E|Y_n| <\infty$.

(1): $\langle EX, e_n \rangle =E\langle X, e_n \rangle$ (since Bochner integral coincides with Pettis integral). Hence $EX =\sum \langle EX, e_n \rangle e_n =\sum E\langle X, e_n \rangle e_n$.

Related Question