Chain rule for already conditioned probability

47 Views Asked by At

I know that the chain rule says:

$P(A_1, A_2, A_3) = P(A_1 | A_2, A_3) P(A_2 | A_3) P(A_3)$

Is it then correct to say that

$P(A_1, A_2, A_3 | A_4) = P(A_1 | A_2, A_3, A_4) P(A_2 | A_3, A_4) P(A_3 | A_4)$

Or should I multiply by $P(A_4)$ at the end?

1

There are 1 best solutions below

0
On BEST ANSWER

You can find the correct form using Bayes' theorem: $P(A_1,A_2,A_3|A_4) = \dfrac{P(A_1,A_2,A_3,A_4)}{P(A_4)}$. So you would need $P(A_1,A_2,A_3,A_4)$ which is equal to $P(A_1 | A_2, A_3,A_4) P(A_2 | A_3,A_4) P(A_3|A_4) P(A_4)$.

This means that $P(A_1,A_2,A_3|A_4) = P(A_1 | A_2, A_3,A_4) P(A_2 | A_3,A_4) P(A_3|A_4)$.