Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

For full disclosure, this problem is drawn from my latest problem set, but at this point, I care more about solving the puzzle than about the credit. The statement is:

Given iid continuous random variables $X_1,X_2,X_3$, compute $P(X_1 > X_2 | X_1 > X_3)$.

Obviously, that's equal to $\frac{P(X_1 > X_2 ,\ X_1>X_3)}{P(X_1>X_3)}$ and that the denominator is $\frac{1}{2}$ is clear enough, but I can't quite grep the numerator. Class discussion and my own work lead me to believe it will be a triple integral on the general joint probability function (separable into the marginals due to independence), but I can't figure out how to set the limits, and then how to get a numeric answer from there. Any hints in the right direction would be greatly appreciated.

share|improve this question
add comment

2 Answers 2

up vote 1 down vote accepted

Method 1: The numerator should be $1/3$. HINT: What should be the probability that the maximum of the three numbers $\{ X_1, X_2, X_3 \}$ turns out to be $X_1$?


Method 2: Getting the same answer using integrals: Let $F(\cdot)$ be the c.d.f. of the distribution. Conditioned on $X_1 = x$, the probability that $X_2$ (resp. $X_3$) lies in the interval $(-\infty, x)$ is $F(x)$. Therefore, the desired probability is $$ p = \int_{-\infty}^{+\infty} F(x)^2 F^{\ \prime}(x) dx. $$ Using integration by parts, $$ p = \left. F(x)^2 F(x) \right|_{-\infty}^{+\infty} - \int_{-\infty}^{\infty} 2 F(x) F^{\ \prime}(x) \times F(x) dx = \left. F(x)^3 \right|_{-\infty}^{+\infty} - 2p. $$ Can you now simplify to get $p = 1/3$?


Method 3: Using triple integrals: Let $f(\cdot)$ denote the p.d.f. of the distribution. Then, the "happy" event whose probability we are computing is that $(X_1, X_2, X_3) \in D$ where $$ D := \{ (x, y, z) \mid x \geq y, x \geq z \} \subseteq \mathbb R^3. $$ So the probability is given by $$ \iiint_D f_{X_1, X_2, X_3}(x,y,z) \ dx \ dy \ dz = \iiint_D f(x) f(y) f(z) \ dx \ dy \ dz. $$ Now, fixing an $x \in \mathbb R$, we note that $y$ and $z$ lie in the interval $(-\infty, x)$. So the probability is given by $$ \int_{-\infty}^\infty f(x) \left( \int_{-\infty}^x f(y) dy \right) \left( \int_{-\infty}^x f(z) dz \right) dx, $$ which can be written equivalently as $$ \int_{-\infty}^\infty f(x) F(x)^2 dx, $$ confirming the integral in Method 2.


Another approach. The original problem can also be solved directly as follows. Sort the three numbers $X_1, X_2, X_3$ and rank them in decreasing order, so that $r_i$ is the rank of $X_i$. (We assume that $X_1, X_2, X_3$ are pairwise distinct; this assumption is ok if this fails to occur with zero probability.) Then since $(X_1, X_2, X_3)$ are iid, the random variable $r := (r_1, r_2, r_3)$ has the uniform distribution over the $3! = 6$ permutations of $(1, 2, 3)$.

Conditioned on $X_1 > X_3$, $r$ takes the uniform distribution over $\{(1,2,3), (1,3,2), (2,1,3)\}$. Of these three possibilities, two of them (namely, ${ (1, 2, 3), (1, 3, 2)}$) correspond to the event $X_1 > X_2$. Hence the probability that $X_1 > X_2$ conditioned on $X_1 > X_3$ is $\frac {2}{3}$.

share|improve this answer
add comment

Let $I=\text{the index }i\text{ for which }X_i=\max\{X_1,X_2,X_3\}$. Then $$\Pr(X_1 > X_2\text{ and }X_1>X_3) = \Pr(I=1) = \frac13$$ simply because all three possible values of $I$ are equally probable, by symmetry.

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.