Stats Problem using Stirling's Approx.

  • Thread starter Thread starter wintermute++
  • Start date Start date
  • Tags Tags
    Stats
AI Thread Summary
The discussion focuses on using Stirling's approximation to simplify the expression for Pn,k, where n and k are large integers. Participants clarify the application of Stirling's formula to both the numerator and denominator of the factorial ratio. There is a debate about whether k is small compared to n, which could allow for further simplification of the expression. The conversation also includes corrections and confirmations regarding the calculations, leading to a consensus on the final approximation. Overall, the thread emphasizes the importance of careful application of Stirling's approximation in statistical problems.
wintermute++
Messages
29
Reaction score
0

Homework Statement


Let n and k be positive integers such that both n and n − k are large. Use Stirling’s formula to write as simple
an approximation as you can for Pn,k.

Homework Equations


## \lim_{n \rightarrow \infty} {(2 \pi)^{1/2}n^{n+1/2}e^{-n} \over n!} = 1 ##

The Attempt at a Solution


Since ## n - k ## is also large, I assume you apply Stirling's approximation to both the numerator and denominator of ## {n! \over (n-k)!} ##.

## {(2 \pi)^{1/2}n^{n+1/2}e^{-n} \over (2 \pi)^{1/2}(n-k)^{n-k+1/2}e^{-n+k}} = {n^{n+1/2} \over (n-k)^{n-k+1/2}e^{k}} ##

I'm teaching myself statistics so I don't have anyone to turn to for help. Does this answer look correct?
 
Physics news on Phys.org
wintermute++ said:
## {(2 \pi)^{1/2}n^{n+1/2}e^{-n} \over (2 \pi)^{1/2}(n-k)^{n-k+1/2}e^{-n+k}} ##
No, you've lost some terms in the substitution. Please post your working to reach the above expression.
 
haruspex said:
No, you've lost some terms in the substitution. Please post your working to reach the above expression.

Step 1. ## n! = (2 \pi)^{1/2}n^{n+1/2}e^{-n} ## and ## (n-k)! = (2 \pi)^{1/2}(n-k)^{n-k+1/2}e^{-n+k} ##

Step 2. ## {n! \over (n-k)!} = {(2 \pi)^{1/2}n^{n+1/2}e^{-n} \over (2 \pi)^{1/2}(n-k)^{n-k+1/2}e^{-n+k}} ##
 
wintermute++ said:
Step 1. ## n! = (2 \pi)^{1/2}n^{n+1/2}e^{-n} ## and ## (n-k)! = (2 \pi)^{1/2}(n-k)^{n-k+1/2}e^{-n+k} ##

Step 2. ## {n! \over (n-k)!} = {(2 \pi)^{1/2}n^{n+1/2}e^{-n} \over (2 \pi)^{1/2}(n-k)^{n-k+1/2}e^{-n+k}} ##
Sorry, I misread it as Cn,k. So I agree with your answer now.
It doesn't say whether k is small compared with n. If it were you could simplify it further.
 
haruspex said:
Sorry, I misread it as Cn,k. So I agree with your answer now.
It doesn't say whether k is small compared with n. If it were you could simplify it further.

If that were the case, it would be approximately 1, correct?
 
wintermute++ said:
If that were the case, it would be approximately 1, correct?
No, I get ##n^k e^{k/2n - k^2/n}##
 
haruspex said:
No, I get ##n^k e^{k/2n - k^2/n}##

Would you mind explaining how you got that answer? I get ## {n^k \over e^k} ##
 
wintermute++ said:
Would you mind explaining how you got that answer? I get ## {n^k \over e^k} ##
##{n^{n+1/2} \over (n-k)^{n-k+1/2}e^{k}}##
##{n^kn^{n-k+1/2} \over (n-k)^{n-k+1/2}e^{k}}##
##{n^k \over (1-k/n)^{n-k+1/2}e^{k}}##
##{n^k \over e^{-(k/n)(n-k+1/2)}e^{k}}##
##{n^k \over e^{-(k/n)(-k+1/2)}}##
##{n^k e^{(k/n)(-k+1/2)}}##
##{n^k e^{-k^2/n+k/2n}}##
I guess we can drop the k/2n.
 
haruspex said:
##{n^{n+1/2} \over (n-k)^{n-k+1/2}e^{k}}##
##{n^kn^{n-k+1/2} \over (n-k)^{n-k+1/2}e^{k}}##
##{n^k \over (1-k/n)^{n-k+1/2}e^{k}}##
##{n^k \over e^{-(k/n)(n-k+1/2)}e^{k}}##
##{n^k \over e^{-(k/n)(-k+1/2)}}##
##{n^k e^{(k/n)(-k+1/2)}}##
##{n^k e^{-k^2/n+k/2n}}##
I guess we can drop the k/2n.

I see, very nice. Many thanks for your help earlier too.
 

Similar threads

Back
Top