How to show implication for equivalent equations?

831 Views Asked by At

What does it mean for one event to cause another event — for example, smoking () to cause cancer ()? There is a long history in philosophy, statistics, and the sciences of trying to clearly analyze the concept of a cause. One tradition says that causes raise the probability of their effects; we may write this symbolically is (|)>().(1)

Another way to formulate a probabilistic theory of causation is to say that (|)>(|^).(2)

I know that equation (1) implies that (|)>().

Show that equation (1) implies equation (2)

1

There are 1 best solutions below

0
On BEST ANSWER

By the law of total probability, we can relate $P(E)$ to $P(E\mid C)$ and $P(E\mid C^c)$

$$P(E) = P(E\mid C)P(C) + P(E\mid C^c)(1-P(C)) = P(C)[P(E\mid C) - P(E\mid C^c)] + P(E\mid C^c) $$

Using our assumption $P(E\mid C) > P(E)$ we get:

$$P(C)[P(E\mid C) - P(E\mid C^c)] + P(E\mid C^c) = P(E) < P(E\mid C)$$

Collecting conditional probabilities on either side of the inequality we get

$$ P(E\mid C^c)- P(C)P(E\mid C^c) < P(E\mid C) - P(C)P(E\mid C) $$

Finally, we factor out $1-P(C)$ on both sides:

$$ (1-P(C))P(E\mid C^c) < (1-P(C))P(E\mid C)$$

Assuming $P(C) < 1$ ($P(C)=1$ would make it omnipresent and not a typical "cause"), we can divide by the strictly positive value $1-P(C)$ to get our implication:

$$P(C) < 1 \text{ and } (1-P(C))P(E\mid C^c) < (1-P(C))P(E\mid C) \implies P(E\mid C^c) < P(E\mid C)\;\; \square$$