Given events $A$, $B$ and $C$, I am somewhat lost as to how to compute exactly one of them happens. I was thinking of defining the event where you pass exactly one of them as so:

$D$ = event exactly one of the three events occur

$$D = (A \cap B'\cap C')\cup(A'\cap B\cap C')\cup (A'\cap B'\cap C) = A1\cup A2\cup A3$$

Now, since they're mutually independent, I understand $P(A'\cap B'\cap C)$, for example, can become $P(A')P(B')P(C)$. But since they aren't mutually exclusive, I cannot split up $P(D) = P(A1) + P(A2) + P(A3)$.

I drew a venn diagram and saw that the regions of interest $A1$, $A2$ and $A3$ are indeed disjoint, but I am not sure if this applies for all possible $A1$, $A2$ and $A3$, and so I am hesitant to assume this until I can be sure these events are disjoint.


$A1= A\cap B' \cap C'$ and $A2= A'\cap B \cap C'$ are mutually exclusive since for $A \cap B' \cap C'$ to happen we need $A$ to happen, and for the second one we need $A$ to not happen. So, $A1$ and $A2$ are mutually exclusive. Likewise $A2$ and $A3$ are mutually exclusive, and so are $A1$ and $A3$. So, they are all mutually exclusive.

This can be generalized as follows:

Say we have a numbers of events $B_1$ through $B_n$. Let $A_i$ be the event that $B_i$ happens but none of $B_j$ with $j \not = i$. Then any $A_i$ and $A_j$ with $i \not = j$ are mutually exclusive, since for $A_i$ to happen you need $B_i$ to happen, but for $A_j$ to happen you need $B_j$ to not happen. So, all the $A_i$'s are mutually exclusive.


Although @Bram28's answer is great, I want to add an answer because I have a resource on an algorithm written specifically for this. Here is a fast recursive algorithm to calculate the exact probability of $m$ out of $n$ events happening. It is important to note that this algorithm calculates the probability for each value of $m$. I will use the notation $P_k$ to denote the probability of the $k^{th}$ event and $P_{k,J}$ to denote the probability of $k$ out of $J$ events happening.

  1. Give your events a numbered order. The order itself does not matter, but you have to give a consistent order.
  2. Assuming you know your probabilities, use the following recursion formulas to get $P_{k,J}$.
    1. $P_{0,0}=1$
    2. $P_{0,J+1}=P_{0,J}\cdot(1-P_{J+1})$
    3. $P_{k+1,J+1}=P_{k,J}\cdot P_{k+1}$
    4. $P_{k,J+1}=P_{k,J}\cdot(1-P_{k+1})+P_{k-1,J}\cdot P_{k+1}$

You should use the recursive formulas first, then use the fourth one to fill the gaps left behind by the other ones.


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.