Poisson distribution
In a minute
A counts events in a fixed stretch of time or space, when they happen independently at a steady average rate. Its mean and variance are both , your GDC gives its probabilities, and independent Poisson counts add to another Poisson count.
You'll be able to
- Decide whether a Poisson model fits, write , and state its assumptions in context.
- Use the mean and variance of , both equal to , without proving them, including to judge whether data look Poisson.
- Find Poisson probabilities on your GDC, after scaling the mean to the interval.
- Add independent Poisson counts by adding their means.
- Choose between normal, binomial and Poisson models, and combine them in multi-step questions.
Before you start3 quick questions on what this builds on
Revise first: 4.6 Probability diagrams, rules and independence, 4.7 Discrete random variables and expected value, 4.8 Binomial distribution, 4.9 Normal distribution and 4.14 Linear combinations and unbiased estimates
When a Poisson model fits
How many orders reach a website in an hour? How many potholes are on a kilometre of road? Each counts events in a fixed interval of time or space. There is no fixed number of trials, so the count has no upper limit.
Independent means one event doesn't make another more or less likely. Uniform average rate means the rate stays the same throughout the interval you care about: a café's lunchtime rush and quiet afternoon have different rates.
is a , and the single number fixes its whole distribution. It is the mean, and it is also the . Both results are in the formula booklet, and you don't need to prove them.
Try it first, then reveal one step at a time.
- (a) counts events in one hour, with mean . So .
- Assumption 1: customers order independently of each other. A promotional email that makes many people order at once would break this.
- Assumption 2: orders arrive at a uniform average rate during the hour. Across a whole day this is doubtful, as evenings may be busier.
- (b) , so the standard deviation is orders.
With data, compare the sample mean with the of the variance, , from 4.14. If they are close, a Poisson model is worth trying. If they are far apart, it probably isn't suitable.
Try it first, then reveal one step at a time.
- (a) Enter the errors as data and the pages as frequencies, then run one-variable statistics. Check that .
- It gives and Square the unrounded value: .
- (b) A Poisson distribution has mean equal to variance, and and are very close. So is worth trying, if errors happen independently at a steady rate across pages.
Poisson probabilities on your GDC
Poisson pdf gives . Poisson cdf gives . Both need only the mean .
takes only whole numbers, so the rules from 4.8 still work: "fewer than " is , and .
Try it first, then reveal one step at a time.
- Let be the number of errors in a day, so . You want .
- Poisson pdf with mean and gives .
- Interpret: on about of days the server logs exactly errors.
Because the rate is uniform, the mean grows in proportion to the interval. A mean of per hour becomes in hours.
Try it first, then reveal one step at a time.
- (a) One hour is minutes, so the mean is , and .
- . Poisson cdf with mean and upper bound gives
- So .
- (b) The model assumes a uniform average rate. Some times of day may be busier, so one hour's rate may differ from the average.
Adding independent Poisson counts
A help-desk inbox gets emails from staff and from customers. If each stream arrives independently at a steady rate, so does the combined stream, at the two rates added together.
This is also why the mean scales with the interval: two independent hours, each , add to .
Try it first, then reveal one step at a time.
- Let and . They are independent, so .
- "More than " is : .
- Poisson cdf with mean and upper bound gives , so .
Try it first, then reveal one step at a time.
- In one hour the total is .
- For hours the mean is , so .
- "Fewer than " is . Poisson cdf gives .
- Check: is below the mean , so a probability under makes sense.
Normal, binomial or Poisson?
Exam questions often leave you to choose the model. Ask what the random variable is.
1. A continuous measurement, such as a mass or time, with a symmetric bell-shaped spread: a , from 4.9.
2. Successes in a fixed number of independent trials with the same probability : a . The count can't exceed .
3. Events in an interval of time or space, independent and at a uniform average rate, with no upper limit: a Poisson distribution.
Try it first, then reveal one step at a time.
- (a) A mass is a continuous measurement that varies symmetrically about the target. Normal.
- (b) independent trials, each with . Binomial, .
- (c) Events counted in a fixed time, with no fixed number of trials. If cyclists pass independently at a steady rate, Poisson.
- (a) Normal: a continuous measurement. (b) Binomial: a fixed number of independent trials. (c) Poisson: events counted in a time interval.
Some questions use two models in turn. A Poisson probability for one interval becomes the of a binomial across several independent intervals.
Try it first, then reveal one step at a time.
- For one trip, , and
- Each trip is now a trial with , so the number of trips with a dolphin is .
- Binomial pdf with , the stored and gives .
- Rounding to first gives Here it still rounds to , but in other questions an early-rounded can change the final answer.
Common mistakes
- Using the mean for the wrong interval. Scale to the question's interval first, and write the new distribution down.
- Writing , which drops . Use .
- Saying "the mean and variance differ" to explain why a situation may not be Poisson when no data are given. Give a reason from the context: events that cluster, or a rate that changes.
- Rounding a Poisson probability before using it as in a binomial. Store the full value and round only the final answer.
Exam tips
- Find a Poisson probability: write for the right interval and the probability statement, then the GDC value to 3 s.f.
- Explain why a Poisson model suits a situation: independence or a uniform average rate, in context. From data: the sample mean and variance are about equal.
- A switch from "in an interval" to "on how many of the days" usually means a binomial, with the Poisson probability as .
Key takeaways
Practise this
19 past-paper questions on Poisson distribution in the AI HL bank, each with a mark scheme and a worked explanation.