Content
Summary
The Poisson distribution is a discrete probability model for unbounded count data within a fixed interval, defined by a mean parameter μ where variance equals the mean. It assumes events occur independently at a constant rate, making it suitable for modeling rare events or traffic patterns. The model approximates the binomial distribution when trials are numerous and success probability is low.
Intuition
Think of the Poisson distribution as a tool for counting things that happen randomly but steadily over time or space, like emails arriving or cars passing a checkpoint. Unlike the binomial distribution, which counts successes in a fixed number of tries, Poisson handles situations where the 'number of tries' is effectively infinite or unknown, as long as the average rate is stable. A key feature is that the spread (variance) of the data is exactly equal to the average count, meaning the standard deviation is simply the square root of the mean. This relationship helps us understand that as the average count increases, the distribution becomes more symmetric and bell-shaped.
Key equations
Probability mass function for observing exactly x events with mean μ.
Variance of the Poisson distribution equals the mean.
Standard deviation is the square root of the mean.
Approximate 95% confidence interval for large counts C.
Derivation
- The Poisson distribution arises as a limit of the binomial distribution when the number of trials n is large (>20) and the probability of success p is small (<0.05), such that np remains constant.
- It models events occurring independently at a constant average rate μ within a fixed interval.
- The probability of observing x events is derived from the limit of the binomial formula, resulting in the PMF involving Euler's number e.
- For non-unit intervals, the rate λ is scaled by time t to get μ = tλ.
- For large observed counts C (>25), the distribution approximates a Gaussian, allowing the use of the Normal approximation for confidence intervals.
Worked example
A bakery places an average of 12 loaves of bread on a shelf every 30 minutes. What is the probability that exactly 3 loaves are placed in a 5-minute interval?
- Identify the base rate: 12 loaves per 30 minutes.
- Scale the rate to the interval of interest: 5 minutes is 1/6 of 30 minutes.
- Calculate the new mean μ: 12 * (5/30) = 2 loaves.
- Apply the Poisson PMF for x=3 and μ=2: P(X=3) = (2^3 * e^-2) / 3!.
- Compute the value: (8 * 0.1353) / 6 ≈ 0.1804.
Answer 0.1804
Common mistakes
- Ignoring the discrete nature of the variable when calculating cumulative probabilities, such as failing to include all integers greater than a threshold.
- Applying the distribution to data with varying rates without subdividing the interval into homogeneous segments.
- Using standard printed Poisson tables for mean values μ exceeding 20; electronic tables or Normal approximation should be used instead.
- Calculating confidence intervals using a normalized rate instead of the actual total observed count C.
- Applying the Poisson approximation to a binomial distribution when n is small or p is large.