Re: Need help: Binomial Coefficients Problem



In article <1138673025.660488.197050@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
tutorny <tutorny@xxxxxxxxx> wrote:

> I have the following problem:
>
> I solved the probability of some event to be equal to:
> C(10,2)*C(k-10,18)/C(k,20).
>
> The problem asks to find the value of k for which that probability is
> maximized.
>
> So, I set C(10,2)*C(k-10,18)/C(k,20) = 1.
>
> Can somebody help me solve this problem? I think that some rule of
> binomial coefficients must apply to make this really simple, but can't
> think of which one. The answer is k=100, but I have no idea what
> procedure to follow to get it.

For this to be a probability, we must have k >= 28.

k = 99 is also a solution. But why do you think 1 is the maximum? The
actual maximum is 101355025/318555566, and you're unlikely to get it
without some kind of computer. On the other hand, you can figure out k
= 99 with hardly any calculation.

Remember that C(a,b) = a * (a-1) * (a-2) * ... * (a-b+1)/b! Thus when
you plug in a = k-10 and b = 18 you get a polynomial in k of degree 18:

(k-10)(k-11)....(k-27)
C(k-10,18) =------------------------
18!

and similarly

k(k-1)...(k-19)
C(k,20) =-------------------
20!

Forget about the C(10,2), the 18! and the 20! -- these combine to give
only a constant multiple. When you compute your solution you get a
polynomial of degree 18 divided by a polynomial of degree 20. But
there are some factors in common: k-10, k-11, ... through k-19.
Cancelling these, your probability is of the form

(k-20)(k-21)...(k-27)
const * ---------------------
k(k-1)....(k-9)

i.e. a polynomial of degree 8 over a polynomial of degree 10. If you
set

(k-20)(k-21)...(k-27)
g(k) = ---------------------
k(k-1)....(k-9)

Now the ratio of the probabilities for k+1 and k is exactly

g(k+1) (k-9)(k-19)
------ = --------------
g(k) (k+1)(k-27)

It's easy to find where g(k+1) < g(k) and g(k+1) < g(k): for values
with k >= 28,

(k-9)(k-19) < (k+1)(k-27) <==> -28k + 171 < -27 - 26k <==> 99 < k

while

(k-9)(k-19) > (k+1)(k-27) <==> -28k + 171 > -27 - 26k <==> 99 > k.

Thus g(99) = g(100) and this is the maximum value of g on the integers
>= 28.

In this form you should be able to generalize the problem.

--Ron Bruck
.



Relevant Pages

  • Re: What is Intelligent Design?
    ... probability of given object or 'information system' as they like to ... I have never seen an advocate of ID propose a calculation ... a given object as a product of evolution. ... randomly assembling entirely new genomes from scratch? ...
    (talk.origins)
  • Re: What is Intelligent Design?
    ... probability of given object or 'information system' as they like to ... characteristics of a rock to determine if a rock was chipped for use as ... I have never seen an advocate of ID propose a calculation ... a given object as a product of evolution. ...
    (talk.origins)
  • Error in Allis and Drexler paper
    ... In section 4.4 they compute the probability ... displacement of the tool tip causing errors. ... The calculation effectively compares one displaced ... configurations hugely outnumber the undisplaced. ...
    (sci.nanotech)
  • Re: Coin flips and probability of range of heads
    ... For 10,000 coin flips, the permitted range is 4,900 to 5,100. ... How is this calculation done? ... The formula is the probability mass function, ...
    (sci.stat.math)
  • Re: What is Intelligent Design?
    ... probability of given object or 'information system' as they like to ... characteristics of a rock to determine if a rock was chipped for use as ... I have never seen an advocate of ID propose a calculation ... a given object as a product of evolution. ...
    (talk.origins)