# Does P(A|B,C)P(B)= P(A|C)? [closed]

Can you remove a conditional variable by multiplying by the probability of it?

We know that

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

and this remains true when conditional on $$C$$:

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

You'll notice that the $$C$$ conditioning seems to be "along for the ride."

In Bayesian analysis, one typically replaces $$C$$ with $$I$$, indicating prior information. So,

$$P(A,B|I)=P(A|B,I)P(B|I).$$

which can get cumbersome, but it underscores the central point that every interpretation of probability is conditional on what we -- most often tacitly, but occasionally explicitly -- affirm to be true.

If you want to "remove a conditional," you may use the definition of conditional probability to restate the problem as a joint probability and a marginal: $$P(A|C)=P(A,C)/P(C).$$ Where from there depends on what you are trying to do.

With a little probability algebra, we have:

\begin{aligned} \mathbb{P}(A|B,C) &= \frac{\mathbb{P}(A,B,C)}{\mathbb{P}(B,C)} \\[6pt] &= \frac{\mathbb{P}(C) \cdot \mathbb{P}(A|C) \cdot \mathbb{P}(B|A,C)}{\mathbb{P}(B,C)} \\[6pt] &= \mathbb{P}(A|C) \times\frac{\mathbb{P}(B|A,C)}{\mathbb{P}(B|C)}. \\[6pt] \end{aligned}

Thus, we can see that removal of the conditioning event $$B$$ requires us to multiply by $$\mathbb{P}(B|A,C)/\mathbb{P}(B|C)$$, which is not generally equal to $$\mathbb{P}(B)$$.

Not quite, as $$B$$ may occur or not occur on the right hand side while the left hand side is $$\mathbb P(A,B\mid C)$$

But since $$\mathbb P(A,B\mid C) + \mathbb P(A, B^c\mid C)= \mathbb P(A\mid C)$$, you can say:

$$\mathbb P(A\mid B,C)\mathbb P(B) + \mathbb P(A\mid B^c,C)\mathbb P(B^c)= \mathbb P(A\mid C)$$

It does not work out, since:

$$P(A|B,C)P(B) = \frac{P(A,B,C)}{P(B,C)}P(B) = \frac{P(A,C|B)P(B)}{P(C|B)P(B)}P(B) = P(A|C)P(B)$$

For a more concrete example, take an equally probable six sided die. Let A = a 2 is rolled, B = a 4 is rolled, C = an even number is rolled.

P(A|B,C)P(B) = 0, P(A|C) = 1/3.