Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

We know that by using Stirling approximation: $\log n! \approx n \log n$

So how to approximate $\log {m \choose n}$?

share|cite|improve this question
up vote 18 down vote accepted

A better approximation for the logarithm of a factorial can be found by using $\log n! \approx n \log n - n$. Interestingly, the additional terms in the approximation of the binomial coefficient cancel out, and the result is the same as if you used the simpler approximation $\log n! \approx n\log n$:

$$\begin{align} \log {n\choose m} & = \log \frac{n!}{m!(n-m)!} \\ & = \log n! - \log m! - \log (n-m)! \\ & \approx n \log n - n - m \log m + m - (n-m) \log (n-m) + n-m \\ & = n \log n - m \log m - (n - m) \log (n-m) \end{align}$$

An even better approximation uses more terms of Stirling's approximation, giving $\log n! \approx (n+\tfrac{1}{2})\log n - n + \tfrac{1}{2}\log 2\pi$ and hence

$$\begin{align} \log {n\choose m} &\approx (n+\tfrac{1}{2})\log n - (m+\tfrac{1}{2})\log m - (n-m+\tfrac{1}{2})\log (n-m) - \tfrac{1}{2}\log 2\pi \\ & = n\log n - m \log m - (n-m)\log(n-m) + \cdots \\ & \qquad +\tfrac{1}{2} (\log n - \log m - \log (n-m) - \log 2\pi) \end{align}$$

where the last term is the correction from using more terms of Stirling's approximation.

share|cite|improve this answer

Using $\log(n!) ≈ n \log(n)$ and the definition of the binomial coefficient, $\log{m \choose n} ≈ m \log{m} - (m-n) \log{(m-n)} - n \log{n}$. The same should work for any of the more precise statements of Stirling's approximation.

share|cite|improve this answer

If you expand stirling's approximation for $n\choose\alpha n$, you may also notice that it can be written as $$H(a) - 1/2\log(2\pi\, n\alpha(1-\alpha))$$ Where $$H(a) = \alpha n \log(1/\alpha) -(1 - \alpha) n \log(1/(1 - \alpha))$$ Is the binary entropy function (in nats).

If you further notice that $n\alpha(1-\alpha)$ is the variance of a binomial distribution, it becomes pretty easy to remember. (And you start to see how you might derive the central limit theorem for binomial distributions).

share|cite|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.