Below is the definition of the law of total expectation from Wiki.

enter image description here

The first equation states that for any $X, Y$ on the same probability space, then \begin{equation} E(X) = E(E(X|Y)) \end{equation}

It then states that one special case is

\begin{equation} E(X) = \sum_i E(X|A_i)P(A_i) \end{equation} if $\{A_i\}_i$ is a finite or countable partition of the sample space.

The way that the second equation is prefaced by "one special case.." leads me to think that it is a special case of the first equation.

My question is: how can one relate $E(X) = E(E(X|Y))$ to $E(X) = \sum_i E(X|A_i)P(A_i)$?

That is, if I define $Y = \{A_i\}_{i=1}^{n}$ as the set of partitions on the sample space, is the following correct? $$E(X) = E(E(X|Y)) = \sum_i E(X|A_i)P(A_i)$$


Since the law of total expectation $E(E(X | Y) ) = E(X)$ is more general, it makes sense to show how this implies $E(X) = \sum_{i = 1}^n E(X | A_i)P(A_i)$.

Consider $Y_i = I_{A_i}$ using $\displaystyle \Omega = \dot{\cup}_{i =1}^n A_i$ it follows that $ P \left( \sum_{i =1}^n Y_i = 1 \right) = 1$. Then you have

\begin{align} E(X) & = \sum_{i=1}^n E(X Y_i) \\ & = \sum_{i=1}^n E(X Y_i | Y_i = 1) P(Y_i =1) + E(X Y_i | Y_i = 0) P(Y_i = 0) \\ & = \sum_{i=1}^n E( X | Y_i = 1) P(Y_i =1) + E( 0 |Y_i = 0) P(Y_i = 0) \\ & = \sum_{i=1}^n E( X | A_i) P(A_i). \end{align}

  • $\begingroup$ Thanks. Is it correct to add one last line to the derivation above: i.e., $\sum_{i=1}^n E(X|A_i)P(A_i) = E(E(X|Y))$ $\endgroup$
    – Devon
    Mar 12 at 1:20
  • $\begingroup$ That is just because $E(X) = E(E(X|Y))$. $\endgroup$
    – Manuel
    Mar 12 at 2:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.