0
$\begingroup$

$A_1$, $A_2$, $B_1$ and $B_2$ are random variables with

$P(A_1, A_2, B_1, B_2) = P(A_1, A_2) P(B_1, B_2)$.

Are $A_1$ and $B_1$ independent? I would think not, because if e.g. $P(A_2)=0$ and $P(B_2)=0$, then the equation above would be correct for dependent $A_1$ and $B_1$.

Thanks for any help and thoughts!

$\endgroup$
  • 2
    $\begingroup$ You probably mean events and not random variables. In that case your thinking is correct. $\endgroup$ – Kavi Rama Murthy Nov 15 '18 at 12:07
0
$\begingroup$

As it is said in the comment, be careful not to mix random variables with events : $P(A_1)=0$ means nothing for a random variable. An event is something that might happen or not and it has a probability while a random variable is a variable that has a random value and it has not a single probability.

Events and random variables are related : you can create a random variable from an event by saying that its value is 1 when then event happens and 0 otherwise. Reciprocally, for any random variable $V$, you can consider one event for each of its possible values : $V=1$, $V=2$,...

Your equation (E) : $P(A_1, A_2, B_1, B_2) = P(A_1, A_2) P(B_1, B_2)$ is the definition of independence between the unions of variables $\{A_1,A_2\}$ and $\{B_1,B_2\}$.

But its arguments are random variables and it is a convenient notation that actually means one equation for each combinations of the related events.

eg : $P(A_1=T,A_2=F,B_1=1,B_2=F)=P(A_1=T,A_2=F)P(B_1=T,B_2=F)$

If your random variables are associated with 4 events, (E) is actually 16 equations, not only the one associated with True values.

So if all those 16 equations are true, you have independence of $\{A_1,A_2\}$ and $\{B_1,B_2\}$ that implies the independence between any subsets and so independence between $A_1$ and $B_1$.

$\endgroup$

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.