Conditional variance: $ \operatorname{Var}(\mathbb{E}[X\mid\theta]\mid P) = \operatorname{Var}(X\mid P) - \operatorname{Var}(X\mid \theta)$

56 Views Asked by At

Why does it hold: $ \operatorname{Var}(\mathbb{E}[X\mid\theta]\mid P) = \operatorname{Var}(X\mid P) - \operatorname{Var}(X\mid\theta)$, where $\theta$ are normal random variables? It looks a bit like projection theorem, but I can't apply it here.