Randomly selected numbers on a continuous interval

Calculus Level 5

Ekesh randomly chooses three numbers a , b , a, b, and c c between 0 0 and 1 1 . If the probability P ( max ( a , b , c ) min ( a , b , c ) 1 3 ) P\left(\big| \text{max}(a, b, c) - \text{min}(a, b, c) \big| \geq \frac{1}{3}\right) can be expressed as m n , \frac{m}{n}, where m m and n n are coprime positive integers, what is m + n ? m + n?


The answer is 47.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

1 solution

P ( max ( a , b , c ) min ( a , b , c ) 1 / 3 ) = 1 P ( max ( a , b , c ) ) min ( a , b , c ) 1 / 3 ) = 1 V o l u m e ( R ) P(\max{(a,b,c)}-\min{(a,b,c)} \geq 1/3) = 1 - P(\max{(a,b,c)})-\min{(a,b,c)} \leq 1/3) = 1 - Volume(R) , where R R is the 3d region given by R = { ( a , b , c ) : max ( a , b , c ) ) min ( a , b , c ) 1 / 3 , 0 a , b , c 1 } R = \{(a,b,c) : \max{(a,b,c)})-\min{(a,b,c)} \leq 1/3, 0 \leq a,b,c \leq 1\} . After some thinking you realize that R is the region formed by translating the cube { ( a , b , c ) : 0 a , b , c 1 / 3 } \{(a,b,c) : 0 \leq a,b,c \leq 1/3\} from when its corner at ( 0 , 0 , 0 ) (0,0,0) to when that corner is at ( 2 / 3 , 2 / 3 , 2 / 3 ) (2/3,2/3,2/3) . The volume of this region is the size of this cube ( 1 / 3 ) 3 (1/3)^3 plus the volume of the "cylinder" formed by the translation. The cross-section of this cylinder is a regular hexagon with sidelength 2 / 27 \sqrt{2/27} ; this cross section is translated by 2 / 3 2/\sqrt{3} units. So the overall answer is

1 ( 2 / 3 3 3 / 2 ( 2 / 27 ) 2 + ( 1 / 3 ) 3 ) = 20 / 27 1-(2/\sqrt{3}*3*\sqrt{3}/2*(\sqrt{2/27})^2 + (1/3)^3) = 20/27

Good solution. It's different from my own.

Ekesh Kumar - 2 years, 10 months ago

In general, its trivial to prove that P ( m a x ( x 1 , x 2 , . . . , x n ) m i n ( x 1 , x 2 , . . . , x n ) < p ) = p n + n p ( n 1 ) ( 1 p ) \displaystyle P(max(x_1,x_2,...,x_n)-min(x_1,x_2,...,x_n)<p)=p^n+n\cdot p^{(n-1)}\cdot(1-p) . Substituting p = 1 3 p=\frac{1}{3} and n = 3 n=3 gives the answer, no stereometry required.

Ivo Zerkov - 2 years, 8 months ago

Log in to reply

@Ivo Zerkov - that's a good point, makes for an easier solution. I guess you can do something like done here (http://www.di.fc.ul.pt/~jpn/r/prob/range.html) to show that … but that doesn't seem trivial to me … why is it trivial ? thanks!

Christopher Criscitiello - 2 years, 8 months ago

Log in to reply

If all n n variables end up smaller than p p (happens with probability p n p^n ), the condition is certainly satisfied. Otherwise, the largest variable x k x_k is larger than p p . There’s n n ways to pick x k x_k , 1 p 1-p probability this particular variable satisfies x k > p x_k>p and p n 1 p^{n-1} probability the other n 1 n-1 variables all end up in the interval ( x k p , x k ) (x_k-p,x_k) , which is what the condition equates to.

Adding up both possibilities leads to P ( m a x ( x 1 , x 2 , . . . , x n ) m i n ( x 1 , x 2 , . . . , x n ) < p ) = p n + n ( 1 p ) ( p n 1 ) P(max(x_1,x_2,...,x_n)-min(x_1,x_2,...,x_n)<p)=p^n+n(1-p)(p^{n-1})

Ivo Zerkov - 2 years, 8 months ago

Log in to reply

@Ivo Zerkov awesome, thanks!

Christopher Criscitiello - 2 years, 8 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...