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.

$x,y,z$ are independent and uniform random on $[0,1]$.

I know that $P(x \geq yz) = 3/4$ by triple integrating the triple density function.

However, is there a easier way to get this solution? Without triple integration.

share|improve this question
2  
I don't know, performing $\int_0^1\int_0^1 (1 - yz) \,dy\,dz$ is pretty easy... what sort of method would you consider to be easier? –  Zarrax Aug 22 '12 at 16:50
    
I was trying to avoid triple integration. Can I reason this way: P(x >= yz) = 1 - yz. So we're just double integrating the function 1-yz over the unit square region? Totally not mentioning triple integrals at all... –  Legendre Aug 22 '12 at 16:57
1  
Given that $y = y_0$ and $z = z_0$, the chance that $x > yz$ is $1 - y_0z_0$. So since that $y$ and $z$ are uniformly distributed in $[0,1]$ you end out with the double integral I wrote above... so you don't have to do any triple integrals (but it's somewhat implicit in what I'm saying here). –  Zarrax Aug 22 '12 at 17:01
    
Thank you so much! –  Legendre Aug 22 '12 at 17:07
add comment

3 Answers

up vote 8 down vote accepted

The conditional probability given $z \in [0,1]$, $P(x \ge yz | z)$, is the area of the region of the square $[0,1]^2]$ below the line $y = x/z$. Since the region above the line is a right triangle with sides $1$ and $z$, we have $P(x \ge y z | z) = 1 - z/2$.
Since $E[z] = 1/2$, the unconditional probability is $E[P(x \ge y z | z)] = E[1 - z/2] = 1 - E[z]/2 = 3/4$.

share|improve this answer
    
Nice answer! Thanks! –  Legendre Aug 22 '12 at 17:19
add comment

$$ \mathrm P(X\leqslant YZ\mid Y,Z)=YZ\qquad\quad \mathrm E(YZ)=\mathrm E(Y)\cdot\mathrm E(Z)=1/4$$

share|improve this answer
    
As explained in @Zarrax's comment. –  Did Aug 22 '12 at 19:29
    
Are you saying that $E(YZ) = P(X \leq YZ)$? If so, why? –  Legendre Aug 23 '12 at 15:48
    
Yes I am. Because (1.) $P(X\leqslant x)=x$ for every $x$in $(0,1)$, (2.) $YZ\in(0,1)$ almost surely, (3.) $YZ$ is independent of $X$. –  Did Aug 23 '12 at 16:10
add comment

Here is an alternative way to see it.

If $U$ is uniformly distributed on $(0,1)$ random variable, $X= -\log(U)$ follows the exponential distribution with unit rate. Thus: $$ \mathbb{P}\left( U_1 \geqslant U_2 U_3 \right) = \mathbb{P}\left(-\log(U_1) \leqslant -\log(U_2) -\log(U_3) \right) = \mathbb{P}\left(X_1 \leqslant X_2 + X_3 \right) = \mathbb{P}\left(\frac{X_1}{X_1+X_2+X_3} \leqslant \frac{1}{2} \right) $$ The ratio $\frac{X_1}{X_1+X_2+X_3}$ follows Beta distribution with parameters $(1,2)$ (e.g., see here, or in "Question about order statistic") meaning that $$ \mathbb{P}\left( U_1 \geqslant U_2 U_3 \right) = \int_0^{\frac{1}{2}} \frac{1-x}{\operatorname{B}(1,2)} \mathrm{d} x \stackrel{\operatorname{B}(1,2)=\frac{1}{2}}{=} \int_\frac{1}{2}^1 2 x \mathrm{d} x = 1^2 - \left(\frac{1}{2}\right)^2 = \frac{3}{4} $$

share|improve this answer
1  
The beta(1,2) distribution is also the distribution of the minimum of two independent uniform[0,1] random variables $X$ and $Y$. (This is a standard result in the theory of order statistics.) In order to have $min(X,Y) \le 1/2$ we must have $X \le 1/2$ or $Y \le 1/2$, which obviously has probability 3/4. –  Michael Lugo Aug 22 '12 at 18:37
    
Whoever cast the down vote, would you please explain what you would like improved about the post? –  Sasha Aug 22 '12 at 19:28
1  
Cute, but maybe more complicated than necessary. –  Michael Hardy Aug 22 '12 at 21:54
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.