marginal probability, three variables

1.7k Views Asked by At

I got stuck with this:

P(A|C) = $\sum_{B}$ (A, B|C)

(B should be beneath the sum sign)

I understand marginal probability with two variables. But when it is three or more variables, I just could not make sense of it. What if A intersects C, but it does not intersect with the intersection of B and C? Why sum of B works? Can someone walk me through this? Thanks!

1

There are 1 best solutions below

3
On BEST ANSWER

You should be aware of the definition of conditional probability.  $\mathsf P(A\mid C)=\mathsf P(A,C)/\mathsf P(C)$

For three events, same again: $\mathsf P(A, B\mid C)=\mathsf P(A,B,C)/\mathsf P(C)$

So just apply that and the Law of Total Probability.

$$\begin{align}\mathsf P(A\mid C) ~=~&\dfrac{\mathsf P(A,C)}{\mathsf P(C)} \\[1ex]=~& \dfrac{\sum_{B_k}\mathsf P(A,B_k,C)}{\mathsf P(C)} \\[1ex]=~& \sum_{B_k}\mathsf P(A, B_k\mid C)\end{align}$$

That is all.


As to why the sum works: If $(B_k)$ is a sequence of disjoint events partitioning the sample space -- well, in this case, at least exhaustively covering $C$ -- then: $$\mathsf P(A\cap C) ~=~ \mathsf P(\bigcup_{B_k} A\cap B_k\cap C)~=~\sum_{B_k}\mathsf P(A\cap B_k\cap C)$$