I was reading Introduction to Probability Models 11th Edition and saw this proof of why Poisson Distribution is the approximation of Binomial Distribution when n is large and p is small:

An important property of the Poisson random variable is that it may be used to approximate a binomial random variable when the binomial parameter $n$ is large and $p$ is small. To see this, suppose that $X$ is a binomial random variable with parameters $(n, p),$ and let $\lambda=n p .$ Then $$ \begin{aligned} P\{X=i\} &=\frac{n !}{(n-i) ! i !} p^{i}(1-p)^{n-i} \\ &=\frac{n !}{(n-i) ! i !}\left(\frac{\lambda}{n}\right)^{i}\left(1-\frac{\lambda}{n}\right)^{n-i} \\ &=\frac{n(n-1) \cdots(n-i+1)}{n^{i}} \frac{\lambda^{i}}{i !} \frac{(1-\lambda / n)^{n}}{(1-\lambda / n)^{i}} \end{aligned} $$ Now, for $n$ large and $p$ small $$ \left(1-\frac{\lambda}{n}\right)^{n} \approx e^{-\lambda}, \quad \frac{n(n-1) \cdots(n-i+1)}{n^{i}} \approx 1, \quad\left(1-\frac{\lambda}{n}\right)^{i} \approx 1 $$ Hence, for $n$ large and $p$ small, $$ P\{X=i\} \approx e^{-\lambda} \frac{\lambda^{i}}{i !} $$

I can understand most part of the proof except for this equation:

$\left(1-\frac{\lambda}{n}\right)^{n} \approx e^{-\lambda}$

I really don't remember where it comes from, could anybody explain this to me? Thanks!.

  • $\begingroup$ It's related to the definition of $e$ itself $\endgroup$
    – Dleep
    May 22 '16 at 9:22
  • $\begingroup$ It's notable that Ross' A first course in probability (I checked 9th and 10th editions) misses the commas in the approximations following for $n$ large and $p$ small, which may bring others to this page, as it did myself. $\endgroup$
    – sunspots
    Sep 6 '21 at 2:49

Well, this is a basic fact of the exponential function $e^x$.

One definition of $e$ is the limit $\lim_{n\to\infty}(1+\frac1n)^n$. By a monotonicity argument one can prove $\lim_{x\to\infty}(1+\frac1x)^x=e$ where $x$ now ranges the real numbers.

Also note that $1-\frac1x=\frac{x-1}x=1/\frac x{x-1}=1/(1+\frac1y)=(1+\frac1y)^{-1}$ where $y=x-1$.
So, one has the following: $$\begin{aligned} \lim_{x\to\infty}(1-\frac1x )^x &= \lim_{y\to\infty}(1+\frac1y )^{-(y+1)} \\ &=\lim_{y\to\infty}(1+\frac1y)^{-y}\times\lim_{y\to\infty}(1+\frac1y)^{-1} \\ &=e^{-1}\times1=e^{-1}\,. \end{aligned} $$

From here, assuming $\lambda>0$, $$\begin{aligned} e^{-\lambda}=(e^{-1})^\lambda &= \lim_{x\to\infty}(1-\frac1x)^{\lambda x} &\to{\ z:=\lambda x} \\ &= \lim_{z\to\infty}(1-\frac\lambda z)^z\,. \end{aligned} $$

In consequence, we have this limit for every sequence $z_n\to\infty$ written in place of $z$ and limiting on the natural $n\to\infty$. In particular, this also holds for $z_n=n$.

Note that we had to take the turnaround for arbitrary real numbers instead of integers only because of the exponent $\lambda$.

  • $\begingroup$ Thanks, that's a really clear explanation! $\endgroup$
    – Demonedge
    May 22 '16 at 10:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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