TAOCP 1.2.10 Exercise 15

The Poisson distribution with parameter $\mu$ assigns probabilities p_k = e^{-\mu}\frac{\mu^k}{k!

Section 1.2.10: Analysis of an Algorithm

Exercise 15. [HM23] When the probability that some quantity has the value $k$ is $e^{-\mu}(\mu^k/k!)$, it is said to have the Poisson distribution with mean $\mu$.

a) What is the generating function for this set of probabilities?

b) What are the values of the semi-invariants?

c) Show that as $n\to\infty$ the Poisson distribution with mean $np$ approaches the normal distribution in the sense of exercise 13.

Verified: yes
Solve time: 1m38s


Solution

The Poisson distribution with parameter $\mu$ assigns probabilities

$$ p_k = e^{-\mu}\frac{\mu^k}{k!}, \qquad k \ge 0. $$

a) Generating function

The generating function is

$$ G(z) = \sum_{k \ge 0} p_k z^k = \sum_{k \ge 0} e^{-\mu}\frac{\mu^k}{k!} z^k. $$

Factor $e^{-\mu}$ and rewrite:

$$ G(z) = e^{-\mu} \sum_{k \ge 0} \frac{(\mu z)^k}{k!}. $$

The exponential series gives $\sum_{k \ge 0} (\mu z)^k/k! = e^{\mu z}$, hence

$$ G(z) = e^{-\mu} e^{\mu z} = e^{\mu(z-1)}. $$

Thus

$$ \boxed{G(z) = e^{\mu(z-1)}}. $$

b) Semi-invariants

Define

$$ K(t) = \log G(e^t). $$

From part (a),

$$ G(e^t) = \exp(\mu(e^t - 1)), $$

so

$$ K(t) = \mu(e^t - 1). $$

Expand $e^t$:

$$ e^t - 1 = \sum_{n \ge 1} \frac{t^n}{n!}. $$

Hence

$$ K(t) = \sum_{n \ge 1} \mu \frac{t^n}{n!}. $$

By definition of semi-invariants (cumulants),

$$ K(t) = \sum_{n \ge 1} \kappa_n \frac{t^n}{n!}, $$

so comparison of coefficients yields

$$ \kappa_n = \mu \quad (n \ge 1). $$

Thus

$$ \boxed{\kappa_n = \mu \text{ for all } n \ge 1}. $$

c) Limit to the normal distribution

Consider the Poisson distribution with mean $np$, so $\mu = np$. Its generating function is

$$ G(z) = \exp(np(z-1)). $$

Let

$$ \sigma = \sqrt{np}. $$

From Exercise 13, the required limit is

$$ \lim_{n \to \infty} e^{-it\mu/\sigma} G(e^{it/\sigma}). $$

Substitute $\mu = np$ and $G$:

$$ e^{-it\mu/\sigma} G(e^{it/\sigma}) = \exp!\left(-it\frac{np}{\sqrt{np}} + np(e^{it/\sqrt{np}} - 1)\right). $$

Simplify the linear term:

$$ -it\frac{np}{\sqrt{np}} = -it\sqrt{np}. $$

Hence the exponent equals

$$ np\left(e^{it/\sqrt{np}} - 1 - \frac{it}{\sqrt{np}}\right). $$

Expand $e^{x}$ at $x = it/\sqrt{np}$:

$$ e^x = 1 + x + \frac{x^2}{2} + O(x^3). $$

Therefore

$$ e^x - 1 - x = \frac{x^2}{2} + O(x^3). $$

With $x = it/\sqrt{np}$,

$$ \frac{x^2}{2} = -\frac{t^2}{2np}, \qquad x^3 = O!\left((np)^{-3/2}\right). $$

Multiply by $np$:

$$ np\left(e^x - 1 - x\right) = -\frac{t^2}{2} + O!\left((np)^{-1/2}\right). $$

As $n \to \infty$, the error term tends to $0$, hence the exponent converges to $-t^2/2$. Therefore

$$ \lim_{n \to \infty} e^{-it\mu/\sigma} G(e^{it/\sigma}) = e^{-t^2/2}. $$

This matches the criterion of Exercise 13, so the Poisson distribution with mean $np$ approaches the normal distribution.

This completes the proof. ∎