1
$\begingroup$

I have a common Bayesian Networks example enter image description here

I need to estimate probablity of JohnCalls given MaryCalls. Other words I need to find

$P(JohnCalls = True|MaryCalls=True)$

From product rule, I know that

$P(JohnCalls = True|MaryCalls=True) = \dfrac{P(JohnCalls, True|MaryCalls=True)}{P(MaryCalls=True)} $

But how can I calculate $P(JohnCalls, True|MaryCalls=True)$ I understand, that it should depend on $Alarm$ variable, but I don't know how.

$\endgroup$
1
$\begingroup$

The equation you have immediately after "From the product rule, I know that..." is incorrect. The definition of conditional probability states that for two random variables $X$ and $Y$,

$$ P(X = x | Y= y) = \frac{P(X=x, Y=y)}{P(Y=y)}. $$

So you should have

$$ P(JohnCalls = True|MaryCalls=True) = \dfrac{P(JohnCalls = True, MaryCalls=True)}{P(MaryCalls=True)}. $$

Note that the expression $P(JohnCalls, True|MaryCalls=True)$ doesn't even really make sense.

Now to compute the probabilities in the numerator and denominator, use the graph structure to factor the joint probability over all the random variables, then marginalize out so that you are left with just the random variable(s) that you care about.

$\endgroup$

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.