I came across this question in Probability & Statistics by Rohatgi Exercise 1.6.9.

Let $P(A|B) = P(A|B\cap C)P(C) + P(A|B\cap C^c)P(C^c)$ with $P(A|B\cap C) \neq P(A|B)$.

If all the three events have non-zero probability, prove that $B$ and $C$ are independent.

After several efforts, I couldn't solve this seemingly innocuous looking question. Any idea on how to proceed will be appreciated.


Since ${\bf P}[B], {\bf P}[A\,\vert\,B]>0$ then, using the given equation, note that we can relate two ways of computing $\,{\bf P}[A \cap B]$:

$${\bf P}[A \,\vert\, B \cap C ]\,{\bf P}[ B \cap C ] + {\bf P}[ A \,\vert\, B \cap C^c ]\,{\bf P}[ B \cap C^c ] \\= {\bf P}[A \cap B ] \\= {\bf P}[A \,\vert\, B \cap C ]\,{\bf P}[B]{\bf P}[C] + {\bf P}[ A \,\vert\, B \cap C^c ]\,{\bf P}[B]\,{\bf P}[C^c]\tag{1}$$

$(1)$ holds non-trivially so that we can subtract the r.h.s from the l.h.s, resulting in

$$\bigg({\bf P}[ B \cap C ]-{\bf P}[ B ]\,{\bf P}[ C ]\bigg)\bigg({\bf P}[A \,\vert\, B \cap C ] - {\bf P}[ A \,\vert\, B \cap C^c ]\bigg)=0\tag{2}$$

Above, for the factorisation, we have used the fact that ${\bf P}[C] > 0$. Now, we know that ${\bf P}[A \,\vert\, B \cap C ] \neq {\bf P}[ A \,\vert\, B \cap C^c ]$ since, otherwise, the given equation in the question implies $ {\bf P}[A \,\vert\, B ] = {\bf P}[ A \,\vert\, B \cap C ]$; a contradiction. Hence, equation $(2)$ implies

$${\bf P}[ B \cap C ]={\bf P}[ B ]\,{\bf P}[ C ]\,.\tag{3}$$

All other probabilities of relevant event combinations - such as ${\bf P}[B\cap C^c] = {\bf P}[B]{\bf P}[C^c]$ - follow by using $(3)$ with basic relationships between probabilities of non-null events and sub-events.

  • $\begingroup$ Thanks. Very elegantly explained. $\endgroup$ – user2808118 Dec 13 '15 at 15:24


Proof by contradiction: Assume that events $B$ and $C$ are dependent events.

Therefore we can say that:

$$P(B \cap C) = P(B) + P(C) - P(B \cup C)$$

Now substitute this in your formula above and if the equality doesn't hold, then the latter is true and $B$ and $C$ are independent events.

  • 3
    $\begingroup$ The equality mentioned by you is true whether or not the events are independent. $\endgroup$ – user2808118 Dec 12 '15 at 14:47

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.