For independent geometric random variables $X, Y$, what is $\Pr(\max(X, Y) = i \cap X = Y)$?

39 Views Asked by At

The following question occurred to me while I was solving exercise 2.7(b) of the first edition of Mitzenmacher-Upfal. Let $X, Y$ be independent geometric random variables. Which of the following claims correctly gives the value of $\Pr(\max(X, Y) = i \cap X = Y)$, and why?

Claim 1:

By the definition of conditional probability, we have

$$\Pr(\max(X, Y) = i \cap X = Y) = \Pr(\max(X, Y) = i | X = Y) \Pr(X = Y)$$

If $X = Y$, then the event "$\max(X, Y) = i$" is equivalent to the event "$X = i \cap Y = i$". Therefore,

$$= \Pr(X = i \cap Y = i)\Pr(X = Y)$$

Claim 2:

The event "$\max(X, Y) = i \cap X = Y$" is equivalent to the event "$X = i \cap Y = i$", thus

$$\Pr(\max(X, Y) = i \cap X = Y) = \Pr(X = i \cap Y = i)$$

1

There are 1 best solutions below

7
On BEST ANSWER

The first claim is wrong. $P(\max(X,Y) = i \mid X=Y)$ is actually $$\frac{P(\max(X,Y) = i \cap X=Y)}{P(X=Y)} = \frac{P(X=i \cap Y=i)}{P(X=Y)}.$$ Your mistake was dropping the conditioning on the event $X=Y$.

After this correction, both approaches give the same answer.