Conditional Expectation of an indicator function is a probability.

1.1k Views Asked by At

If $X_{2}$ is a continous random variable then one needs to show that $\nu_A = E(I_A|X_2)$

or alternatively,

$$\nu_A = \operatorname{argmin}\limits_{E(g(X_2) = X_1, E(g(X_2))^2\lneq \infty} E(X_1-g(X_2))^2$$ is a probability on a probability space(A $\in$ Event Space) and hence one needs to verify the three basic properties of a probability on a space.

$1.\nu_{\emptyset}=0$

$2. 0\leq\nu\leq1 \;\;\forall A$

$3.\nu(\uplus A_i)=\sum_i \nu(A_i)$

The first one seems easy to me as it'd follow directly from the properties of indicator function, is it so, or is there some subtle point I'm missing, and I need some help as to how to prove the second and third condition. Would really appreciate some help.

1

There are 1 best solutions below

0
On

The expected value of an indicator is always a probability since the indicator is a discrete random variable taking on the values $0$ and $1$. So, $$ \begin{array}{lcl} E[I_{A}|X_{2}] &=& 0 \cdot P(I_{A}=0 \,| \,X_{2}) + 1 \cdot P(I_{A}=1 \,| \,X_{2}) = P(A|X_{2})\\ \\ &=& P(I_{A}=1 \, | \, X_{2}). \end{array} $$

So, for example, for disjoint $A_{1}$ and $A_{2}$, $$ E[I_{A_{1} \cup A_{2}}|X_{2}] = P(A_{1} \cup A_{2}|X_{2}) = P(A_{1}|X_{2}) + P(A_{2}|X_{2}). $$