What is the Poisson Distribution ?

The Poisson distribution is highly regarded as one of the most versatile and popular distributions utilized in modern statistics. Before diving into the mathematics of the distribution itself, it is essential to first grasp the specific characteristics that define a Poisson experiment.

Poisson Experiments: Defining the Process

A Poisson experiment is a specific type of statistical experiment characterized by occurrences that happen randomly and independently over a continuous interval, such as time, distance, area, or volume. For an observed phenomenon to qualify as a Poisson experiment, it must strictly satisfy a set of four core criteria:

• The number of successes (events) observed within the specified interval must be quantifiable and countable.
• The mean number of successes, denoted by the parameter λ (lambda), that occurs during the specific interval of time or space must be known and constant.
• The occurrence of an event in one interval must be entirely independent of the occurrence of any other event in a disjoint interval.
• The probability of a success occurring is directly proportional to the size or length of the interval being observed.

Consider the classic example of modeling the number of births per hour at a major regional hospital. Suppose this particular facility has an established average of 10 births per hour. This scenario perfectly exemplifies a Poisson experiment, as it adheres to all four critical properties established above.

• Countable Successes: We can explicitly count the number of births that occur during the one-hour interval.
• Known Mean Rate (λ): It is established that the average (mean) rate is 10 births per hour.
• Independent Outcomes: The probability that one mother gives birth during a given hour is statistically independent of the probability of another mother giving birth.
• Proportional Probability: If we increase the observation interval from one hour to two hours, the probability of observing a birth doubles, demonstrating proportionality to the interval size.

Once we confirm that an event process is Poisson, we can leverage the Poisson distribution formula to address sophisticated questions regarding event likelihoods, such as:

• What is the probability that more than 12 births occur in a given hour?
• What is the probability that less than 5 births occur in a given hour?
• What is the likelihood that the number of births falls between 8 and 11 (inclusive) in a given hour?

The Poisson Distribution Formula and Parameters

The Poisson distribution serves as the mathematical function used to calculate the probability of observing exactly k successes within the specified time or space interval, given the known average rate (λ).

If X represents the number of occurrences (successes) and is considered a random variable that follows a Poisson distribution, the probability mass function (PMF) that X equals a specific value k is determined by the following formula:

P(X=k) = λ^k * e^{– λ} / k!

This powerful formula relies on three essential parameters:

• λ (Lambda): This is the single, crucial parameter of the distribution, representing the mean (average) number of successes that occur during the specific interval being analyzed.
• k: This denotes the actual number of successes or events for which we are calculating the probability (where k = 0, 1, 2, 3, …).
• e: This represents Euler’s number (the base of the natural logarithm), which is an irrational constant approximately equal to 2.71828.

To illustrate the application of this formula, suppose a specific smaller facility experiences a lower average rate of λ = 2 births per hour. We can utilize the Poisson formula above to calculate the probability of observing 0, 1, 2, or 3 births in any given hour:

For 0 births (k=0): P(X=0) = 2⁰ * e^{– 2} / 0! = 0.1353

For 1 birth (k=1): P(X=1) = 2¹ * e^{– 2} / 1! = 0.2707

For 2 births (k=2): P(X=2) = 2² * e^{– 2} / 2! = 0.2707

For 3 births (k=3): P(X=3) = 2³ * e^{– 2} / 3! = 0.1805

By calculating these individual probabilities for all possible outcomes, we can construct a simple histogram that visually represents the frequency of outcomes for this specific probability distribution:

Calculating Cumulative Poisson Probabilities

While calculating a single probability—such as the probability of observing exactly 3 births in an hour—is straightforward using the Probability Mass Function (PMF), statisticians frequently need to determine the likelihood of a range of outcomes. This requires calculating cumulative Poisson probabilities.

A cumulative probability involves summing the individual probabilities (P(X=k)) for all values of k up to and including the desired value. For instance, if we want to know the probability that the hospital experiences 1 or fewer births (P(X≤1)) in a given hour, we must sum the probabilities for 0 births and 1 birth.

Using the individual probabilities we derived previously where λ=2, we calculate the cumulative probability P(X≤1) as follows:

P(X≤1) = P(X=0) + P(X=1) = 0.1353 + 0.2707 = 0.406

We can extend this calculation to find the cumulative probability of experiencing k or less births in a given hour:

1. P(X≤0) = P(X=0) = 0.1353
2. P(X≤1) = P(X=0) + P(X=1) = 0.1353 + 0.2707 = 0.406
3. P(X≤2) = P(X=0) + P(X=1) + P(X=2) = 0.1353 + 0.2707 + 0.2707 = 0.6767

By generating these cumulative probabilities for any number of births up to infinity, we can construct a histogram to visualize this resulting cumulative distribution function:

Key Statistical Properties of the Poisson Distribution

The Poisson distribution possesses several unique and mathematically elegant properties that simplify its use in statistical modeling:

The Mean (Expected Value, E[X]) of the distribution is defined solely by the parameter λ.

The Variance (Var[X]) of the distribution is also numerically equal to the parameter λ.

The Standard Deviation of the distribution is calculated as the square root of the parameter λ, denoted as √λ.

This characteristic—where the mean equals the variance—is a hallmark of the Poisson distribution and is critical for hypothesis testing and model validation. If observed data suggests the mean and variance are significantly different, it suggests that the underlying process may not truly follow a Poisson model.

Returning to our hospital example where the average rate λ = 2 births per hour, these properties yield immediate statistical insights:

The expected (mean) number of births we would anticipate in a given hour is λ = 2 births.

The variance in the number of births we would expect is λ = 2 births.

Applying the Poisson Distribution: Practice Problems

To solidify your understanding of how the Poisson distribution functions in practical scenarios, review the following practice problems. These examples demonstrate calculating exact, ‘greater than’, and cumulative probabilities using the Poisson framework.

Note: We will assume the use of a dedicated statistical calculator to determine the answers to these questions, simplifying the complex computations involving factorials and Euler’s number.

Problem 1: Exact Probability Calculation

Question: An e-commerce website averages (λ) 10 successful sales per hour. In a randomly selected hour, what is the exact probability that the site registers precisely 8 sales?

Answer: Using the Poisson Distribution Calculator with λ = 10 and k = 8, we find that P(X=8) = 0.1126.

Problem 2: ‘Greater Than’ Probability Calculation

Question: A certain realtor maintains an average (λ) of 5 sales per month. In a given month, what is the probability that she achieves more than 7 sales (P(X > 7))?

Answer: Using the Poisson Distribution Calculator with λ = 5 and calculating 1 – P(X ≤ 7), we find that P(X>7) = 0.13337.

Problem 3: Cumulative Probability Calculation

Question: A maternity ward averages (λ) 4 births per hour. In a given hour, what is the probability that 4 or less births occur (P(X ≤ 4))?

Answer: This is a direct cumulative probability calculation. Using the Poisson Distribution Calculator with λ = 4 and k = 4, we find that P(X≤4) = 0.62884.

Further Resources and Statistical Software Application

The Poisson distribution is implemented across virtually all statistical software platforms. For those looking to apply these concepts to larger datasets or in professional settings, the following resources provide guidance on utilizing the Poisson model within specific software environments:

These articles explain how to work with the Poisson distribution in different statistical software packages, optimizing computational efficiency and data analysis for real-world applications.