Mark is practicing for his match against Robin Hood. He shoots at a target board which is a unit circle.
For each real value of x between 0 and 1 , suppose that the probability that each shot of Mark's will be ≤ r from the bullseye is r x . Let f ( x , k ) be the expected value of the minimum area of the circle, whose center is the bullseye, and contains his best ⌊ k n ⌋ shots as n → ∞ .
If
m even ∑ k = 2 ∑ ∞ f ( m 2 , k ) = b a π ,
where a and b are coprime positive integers, then give your answer as a + b .
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.
@Mark Hennings Can you check my solution and see if anything is wrong?
Why is it a Geometry problem? I don't see.
Log in to reply
Two problems. One minor, one major. The coefficient ( t n ) in the definition of the PDF is wrong. You don't just have to choose the t largest, you also have to choose which of the remaining n − t is the next biggest. You should check to see that your PDF does not integrate to 1 .
This is not a major problem, since the way you calculate the expectation cancels this error out.
Much more important, in the very last identity of you calculation of the expectation, you replace n − t by t = n / k twice. For example Γ ( n − t ) becomes Γ ( n / k ) in the denominator. This has the effect of reversing everything so that your formula in the end is what we get if we include the best n / k , as was stated in the original problem, and as I just corrected it.
We should go back to the previous version of the problem, which is consistent with my solution.
Log in to reply
How can I improve the solution? I don't understand why we have to choose which of the remaining ( n − t ) is the next biggest.
Log in to reply
@Kartik Sharma – The simplest way would be to remove ( t n ) from your PDF definitions, and replace it by a constant k . You could then either determine k by checking that the PDF integrates to 1 , or else you could proceed with your expectation calculation, during which the constant k will cancel out.
Or think of it this way. You need to choose which t to exclude, and then you need to decide which of the remaining n − t will be the largest, giving the PDF ( t n ) × ( n − t ) × ( r x ) n − t − 1 × x r x − 1 × ( 1 − r x ) t Note that the coefficient is then t ! ( n − t − 1 ) ! n ! = Γ ( t + 1 ) Γ ( n − t ) Γ ( n + 1 ) = B ( t + 1 , n − t ) 1 which is what is required.
Of course, you will need to rewrite to make it clear that you are either excluding the worst n − ⌊ k n ⌋ , or (equivalently) only including the smallest ⌊ k n ⌋ . You will need to change what you mean by t from what is currently in your proof.
Here is another reason why things must be wrong as they are stated currently.
At the moment, you ask to exclude the top k th of the shots. As you increase k , the proportion of shots that are included therefore increases, and therefore the expected area enclosing them will also increase. That expectation cannot therefore converge to 0 , which would be necessary for convergence. If fact, the limit of the expectation would be π k x 2 ( k − 1 ) x 2 , which tends to π as k → ∞ .
At least, if we are including only the smallest k th of the shots, then the proportion that is included decreases as k increases, and hence it is not unreasonable that the enclosing area should tend to 0 .
As to Geometry, there is not really a good category for questions about continuous random variables. The question would probably best sit in Calculus.
Let r be the minimum radius containing his shots including t = ⌊ k n ⌋ best shots.
Our aim will be to find the probability that the worst among the best t shots lie within r + d r and r .
We select which t shots are best and then which among those best shot is the one lying between r and r + d r .
P(radius containing best t shots is r) = ( t n ) ( 1 t ) ( ( r + d r ) x − r x ) r x ( t − 1 ) ( 1 − r x ) n − t
P(radius containing best t shots is r) = ( t n ) t x r x t − 1 ( 1 − r x ) n − t
Expected Area = ∫ 0 1 r t x − 1 ( 1 − r x ) n − t d t π ∫ 0 1 r t x + 1 ( 1 − r x ) n − t d t
= ∫ 0 1 u t − 1 ( 1 − u ) n − t d t π ∫ 0 1 u t − 1 + 2 / x ( 1 − u ) n − t d t
= Γ ( n + x 2 + 1 ) π Γ ( t + x 2 ) Γ ( n − t + 1 ) Γ ( t ) Γ ( n − t + 1 ) Γ ( n + 1 )
= Γ ( n + x 2 + 1 ) Γ ( ⌊ k n ⌋ ) π Γ ( ⌊ k n ⌋ + x 2 ) Γ ( n + 1 )
We have to find this expected area as n → ∞ .
Now we invoke this (something I recently discovered)-
n → ∞ lim Γ ( m q + n + 1 ) Γ ( m p + t n + 1 ) Γ ( m p + n + 1 ) Γ ( m q + t n + 1 ) = t ∣ p − q ∣ / m 1
Therefore,
n → ∞ lim ( Expected Area ) = f ( x , k ) = k 2 / x π
f ( m 2 , x ) = k m π
k ≥ 2 ∑ m = 1 ∑ ∞ k 2 m π = k ≥ 2 ∑ k 2 − 1 π
which telescopes to
4 3 π
See my comments above.
This needs either to be corrected or deleted, don't you agree?
Log in to reply
I have corrected it now. Thanks, I finally understand it completely. Pls check if anything is wrong.
Log in to reply
The first sentence neglects the worst n / k , instead of only including the best n / k . Otherwise, fine.
Problem Loading...
Note Loading...
Set Loading...
Let R be the distance from the centre of the N th best shot out of N k shots. Then R has PDF g ( r ) = N k × ( N − 1 N k − 1 ) ( r x ) N − 1 × x r x − 1 × ( 1 − r x ) N ( k − 1 ) = B ( N , N ( k − 1 ) + 1 ) x r x − 1 × r x ( N − 1 ) ( 1 − r x ) N ( k − 1 ) 0 ≤ r ≤ 1 and so f ( x , k ) is the limit of the integral π ∫ 0 1 r 2 g ( r ) d r = B ( N , N ( k − 1 ) + 1 ) B ( N + x 2 , N ( k − 1 ) + 1 ) π as N → ∞ . When x = m 2 , where m is a positive integer, we deduce that f ( m 2 , k ) = k m π and since m = 1 ∑ ∞ k 2 m π = k 2 ( 1 − k − 2 ) π = k 2 − 1 π = 2 1 π ( k − 1 1 − k + 1 1 ) we deduce that m e v e n m ≥ 1 ∑ k ≥ 2 ∑ f ( m 2 , k ) = 2 1 π k = 2 ∑ ∞ ( k − 1 1 − k + 1 1 ) = 4 3 π making the answer 7 .