B2.12

Probability and Measure
Part II, 2003

Let HH be a Hilbert space and let VV be a closed subspace of HH. Let xHx \in H. Show that there is a unique decomposition x=u+vx=u+v such that vVv \in V and uVu \in V^{\perp}.

Now suppose (Ω,F,P)(\Omega, \mathcal{F}, \mathbb{P}) is a probability space and let XL2(Ω,F,P)X \in L^{2}(\Omega, \mathcal{F}, \mathbb{P}). Suppose G\mathcal{G} is a sub- σ\sigma-algebra of F\mathcal{F}. Define E(XG)\mathbb{E}(X \mid \mathcal{G}) using a decomposition of the above type. Show that E(E(XG).1A)=E(X.1A)\mathbb{E}\left(\mathbb{E}(X \mid \mathcal{G}) .1_{A}\right)=\mathbb{E}\left(X .1_{A}\right) for each set AGA \in \mathcal{G}.

Let G1G2\mathcal{G}_{1} \subseteq \mathcal{G}_{2} be two sub- σ\sigma-algebras of F\mathcal{F}. Show that (a) E(E(XG1)G2)=E(XG1)\mathbb{E}\left(\mathbb{E}\left(X \mid \mathcal{G}_{1}\right) \mid \mathcal{G}_{2}\right)=\mathbb{E}\left(X \mid \mathcal{G}_{1}\right); (b) E(E(XG2)G1)=E(XG1)\mathbb{E}\left(\mathbb{E}\left(X \mid \mathcal{G}_{2}\right) \mid \mathcal{G}_{1}\right)=\mathbb{E}\left(X \mid \mathcal{G}_{1}\right).

No general theorems about projections on Hilbert spaces may be quoted without proof.