9
$\begingroup$

Below is the scheme of conditional dependence and the probabilities of events:

enter image description here

P(A=1) = 0.01
P(A=0) = 0.99
P(B=1) = 0.1
P(B=0) = 0.9
P(C=1|A=0,B=0) = 0.1
P(C=1|A=0,B=1) = 0.5
P(C=1|A=1,B=0) = 0.6
P(C=1|A=1,B=1) = 0.9

Given the probabilities above I wanted to calculate P(B=1|C=1) and P(B=1|C=1,A=1) but didn't get the correct result.

I wrote the probabilistic function the following way:

P(A, B, C) = P(A)P(B)P(C|A, B)

and then set the variables

P(B=1, C=1) = P(A=0, B=1, C=1) + P(A=1, B=1, C=1)=
=P(A=0)P(B=1)P(C=1|A=0, B=1) + P(A=1)P(B=1)P(C=1|A=1, B=1)=
=0.99*0.1*0.5 + 0.01*0.1*0.9 = 0.0495

The result however is not correct and don't know where is the error. I would be very thankful if anyone could correct/explain what's wrong.

$\endgroup$
4
  • $\begingroup$ You said you were asked P(B=1|C=1) but you computed P(B=1,C=1) hence you are off by the factor P(C=1) (which you will need to compute). $\endgroup$
    – Did
    Mar 30 '12 at 11:06
  • 1
    $\begingroup$ Is P(C=1) = ( P(C=1|A=0,B=0)+P(C=1|A=0,B=1)+P(C=1|A=1,B=0)+P(C=1|A=1,B=1) )/4 ? $\endgroup$ Apr 1 '12 at 11:53
  • $\begingroup$ Maybe because 0.99*0.1*0.5 + 0.01*0.1*0.9 = 0.0504 (and not 0.0495) ? $\endgroup$ Oct 9 '16 at 12:10
  • $\begingroup$ math.stackexchange.com/questions/884728/… $\endgroup$ Apr 1 '17 at 7:03
6
$\begingroup$

The typical way I do inter-causal reasoning is to flip the conditional probabilities around --

$$ \begin{align} P(B = 1 \vert C = 1) & = \frac{P(B = 1, C = 1)}{P(C = 1)} \\ & = \frac{P(C = 1 \vert B = 1)P(B = 1)}{P(C = 1)} \\ \\ P(B = 1 \vert C = 1, A = 1) & = \frac{P(B = 1, C = 1, A = 1)}{P(C = 1, A = 1)} \\ & = \frac{P(C = 1 \vert B = 1, A = 1)P(B = 1)P(A = 1)}{P(C = 1, A = 1)} \\ & = \frac{P(C = 1 \vert B = 1, A = 1) P(B = 1) P(A = 1)}{P(C = 1 \vert A = 1)P(A = 1)} \\ & = \frac{P(C = 1 \vert B = 1, A = 1) P(B = 1)}{P(C = 1 \vert A = 1)} \end{align} $$

Does that help?

$\endgroup$
3
  • 1
    $\begingroup$ And how do you calculate P(C=1) ? Is it ( P(C=1|A=0,B=0) + P(C=1|A=0,B=1) ) * ( P(C=1|A=1,B=0) + P(C=1|A=1,B=1) ) ? $\endgroup$
    – jaor
    Jul 30 '14 at 13:02
  • $\begingroup$ No. That would give you P(C) > 1 $\endgroup$ Oct 10 '16 at 10:04
  • $\begingroup$ This is very old, but how is it that $P(B = 1, C = 1, A = 1) = P(C = 1 \vert B = 1, A = 1)P(B = 1)P(A = 1)$ in the second half? More specifically, why are we assuming that $A$ and $B$ are independent, when it's the case that variable $C$ is observed? As far as I know, in a $v$-structure, the middle variable being observed now means that the two parents are dependent. $\endgroup$
    – Sean
    Dec 12 '20 at 12:58

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.