Skip to content
Tutorials Center
My learning

Topics

Lesson 22 of 57 ยท Common Distributions

The Poisson Distribution

Understand the Poisson distribution for counts of events in a fixed interval: the formula, mean equals variance, the link to the binomial, and worked Python examples.

How many emails arrive in your inbox between 9:00 and 9:10? How many typos are on a page, how many cars pass a toll booth in a minute, how many servers fail in a month? These are counts of events in a fixed interval of time or space, where events happen independently at a steady average rate. The Poisson distribution is the standard model for them.

Unlike the binomial, there is no fixed number of trials and no upper limit on the count. Everything is controlled by a single number, the average rate.

The formula

If events occur independently at a constant average rate, and lambda (the Greek letter ฮป) is the average number of events in the interval you are looking at, then the count X ~ Poisson(lambda) has:

TEXT
P(X = k) = e^(-lambda) * lambda^k / k!,   k = 0, 1, 2, ...

The mean and the variance are both equal to lambda:

TEXT
E[X] = lambda,   Var(X) = lambda

"Variance equals the mean" is a handy check on real data. If your counts vary much more than their average, the Poisson model is probably too simple (the data are overdispersed).

Python Poisson probabilities

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Scaling the rate

The rate must match the interval. If a help desk gets 12 calls per hour on average, then in a 15-minute window lambda = 12 * 0.25 = 3, and in eight hours lambda = 96. A common source of mistakes is to plug the hourly rate into a question about a different window.

Python A help desk

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

When is a process Poisson?

Think of events scattered over time. The count is Poisson when:

  • events happen one at a time (two at exactly the same instant are impossible),
  • the average rate is constant over the interval,
  • the number of events in one interval is independent of the number in any non-overlapping interval.

Rush hours break the constant-rate condition; a failure that triggers many follow-up failures breaks independence. In those cases split the data into calmer periods or use a richer model.

Poisson as a limit of the binomial

Take a binomial with very many trials and a very small success probability, for example 10,000 packets each lost with probability 0.0003. Then n p = 3, and Binomial(n, p) is almost exactly Poisson(3). This is why the Poisson shows up for rare events: it approximates the binomial when n is large and p small, with lambda = n p.

Python Binomial versus Poisson

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Adding independent Poisson counts

If X ~ Poisson(a) and Y ~ Poisson(b) are independent, then X + Y ~ Poisson(a + b). Two independent call centers receiving 3 and 5 calls per hour behave together like one receiving 8 per hour. This addition property makes rate arithmetic safe.

flowchart TD
    A["Counting events in an interval"] --> B{"Fixed number of trials?"}
    B -- "Yes, with success probability p" --> C["Binomial(n, p)"]
    B -- "No, events at a steady rate" --> D["Poisson(lambda)"]
    C -. "n large, p small" .-> D

Choosing between the two counting models.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. A shop averages 2 customers per 5 minutes. What is lambda for a 20-minute window?

    20 minutes is 4 times as long as 5 minutes, so lambda = 2 * 4 = 8.
  2. 2. For a Poisson distribution, the variance equals:

    Both mean and variance of Poisson(lambda) equal lambda.
  3. 3. When does Poisson(n p) approximate Binomial(n, p) well?

    The approximation is good for many trials with a small success probability.
  4. 4. X ~ Poisson(3) and Y ~ Poisson(2) are independent. What is the distribution of X + Y?

    Independent Poisson counts add their rates: 3 + 2 = 5.

Summary

  • The Poisson distribution counts events in a fixed interval when events are independent at a constant rate.
  • P(X = k) = e^(-lambda) lambda^k / k!, with mean and variance both lambda.
  • Always rescale lambda to the interval in the question.
  • It approximates Binomial(n, p) for large n and small p.
  • Independent Poisson counts add their rates.

Next: The Uniform Distribution