How Does the Distribution of Heads in Coin Flipping Change with Large N?

Click For Summary
SUMMARY

The distribution of heads in a large number of coin flips (N) is centered around N/2, with the peak height increasing as N grows. The discussion highlights the importance of using Stirling's approximation and logarithmic expansions to derive the distribution's behavior near the peak. Corrections to earlier calculations reveal that the final expression for the distribution near the peak is -2x²/N, rather than 4x²/N. Additionally, while the peak narrows with increasing N, the width of the peak is inversely proportional to the square root of N.

PREREQUISITES
  • Understanding of probability distributions, specifically binomial distributions.
  • Familiarity with Stirling's approximation for factorials.
  • Basic knowledge of logarithmic functions and their properties.
  • Experience with Taylor series expansions for approximating functions.
NEXT STEPS
  • Study the implications of Stirling's approximation in statistical mechanics.
  • Learn about the Central Limit Theorem and its relation to coin flipping distributions.
  • Explore advanced topics in probability theory, such as convergence of distributions.
  • Investigate the behavior of binomial distributions as N approaches infinity.
USEFUL FOR

Mathematicians, statisticians, and students studying probability theory, particularly those interested in the behavior of distributions in large sample sizes.

chingcx
Messages
21
Reaction score
0

Homework Statement



Flip N fair coins. The distribution for different numbers of heads and tails should be peaked at N/2. When N is very large, the peak will be very high. Let x = N(head)-N/2, required to find an expression for this distribution near the peak, i.e. x<<N.

Homework Equations



Strling Approx
ln(1+x)=x for small x

The Attempt at a Solution



30bjvyr.jpg


clearly incorrect because square of (N/2)! must be smaller than the product of (N/2+x)! and (N/2-x)!, but I can't find where goes wrong.
 
Physics news on Phys.org
1. Going over to the 5th line, you omitted -x \ln{(\frac{N}{2}+x)} and x \ln{(\frac{N}{2}-x)}. These give rise to additional corrections.
-x \ln{(\frac{N}{2}+x)}+x\ln{(\frac{N}{2}-x)}=-x\ln{\left(\frac{1+2x/N}{1-2x/N}\right)}\approx -x\ln{(1+4x/N)}\approx -4x^2 /N

2. In the 7th line, you need to be more careful when expanding to the 2nd order. You should proceed as below.
\ln \left(\frac{N/2}{N/2+x}\right) = \ln \left(\frac{1}{1+2x/N}\right) \approx \ln \left( 1-\frac{2x}{N}+\frac{4x^2}{N^2} \right) \approx -\frac{2x}{N}+\frac{4x^2}{N^2}-\frac{1}{2} {\left(-\frac{2x}{N}+\frac{4x^2}{N^2} \right)}^{2}

\approx -\frac{2x}{N}+\frac{4x^2}{N^2}-\frac{1}{2} \frac{4x^2}{N^2} = -\frac{2x}{N}+\frac{2x^2}{N^2}

and similarly

\ln \left(\frac{N/2}{N/2-x}\right) \approx \frac{2x}{N}+\frac{2x^2}{N^2}


Then, the final answer becomes -2x^2 /N rather than 4x^2 /N.
 
weejee said:
1. Going over to the 5th line, you omitted -x \ln{(\frac{N}{2}+x)} and x \ln{(\frac{N}{2}-x)}. These give rise to additional corrections.
-x \ln{(\frac{N}{2}+x)}+x\ln{(\frac{N}{2}-x)}=-x\ln{\left(\frac{1+2x/N}{1-2x/N}\right)}\approx -x\ln{(1+4x/N)}\approx -4x^2 /N

2. In the 7th line, you need to be more careful when expanding to the 2nd order. You should proceed as below.
\ln \left(\frac{N/2}{N/2+x}\right) = \ln \left(\frac{1}{1+2x/N}\right) \approx \ln \left( 1-\frac{2x}{N}+\frac{4x^2}{N^2} \right) \approx -\frac{2x}{N}+\frac{4x^2}{N^2}-\frac{1}{2} {\left(-\frac{2x}{N}+\frac{4x^2}{N^2} \right)}^{2}

\approx -\frac{2x}{N}+\frac{4x^2}{N^2}-\frac{1}{2} \frac{4x^2}{N^2} = -\frac{2x}{N}+\frac{2x^2}{N^2}

and similarly

\ln \left(\frac{N/2}{N/2-x}\right) \approx \frac{2x}{N}+\frac{2x^2}{N^2}


Then, the final answer becomes -2x^2 /N rather than 4x^2 /N.
Thanks, but it doesn't make sense... when N is very large, the peak should be very narrow, but from the equation, N becomes larger then the exponential decreases much slower.
 
You are right. The width of the peak actually increases, as sqrt(N).
Yet, if you consider (# of heads)/(# of coins), the width decreases as 1/sqrt(N).
 

Similar threads

Replies
4
Views
775
  • · Replies 21 ·
Replies
21
Views
3K
  • · Replies 10 ·
Replies
10
Views
2K
  • · Replies 15 ·
Replies
15
Views
6K
  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 4 ·
Replies
4
Views
3K
Replies
4
Views
4K
  • · Replies 6 ·
Replies
6
Views
4K
  • · Replies 2 ·
Replies
2
Views
2K
Replies
2
Views
2K