Calculating Individual Poisson Probabilities

What an Individual Poisson Probability Answers

A count question often asks for the probability of one specific number of events in a defined window: exactly three calls in ten minutes, exactly two defects on a panel, or exactly one rare event in a batch. The Poisson model gives a direct formula for each such probability. The central details to identify are the count $k$, the observation window, and the parameter $\lambda$ for that same window.

This tutorial focuses on individual probabilities such as $P(X=3)$. That is different from adding probabilities across several counts, as in “three or fewer,” or summing counts above a threshold. Those are cumulative or tail questions.

The Poisson Model and Its Probability Formula

Before calculating, state the model. Let $X$ be the number of occurrences in one fixed observation window or exposure unit. A Poisson model is appropriate when counts are nonnegative whole numbers and the event process is reasonably described by independent occurrences at a stable average rate across the specified window. The window must be defined consistently: for example, one ten-minute interval, one panel, or one batch. For the calculations below, assume that the stated Poisson model and its parameter are appropriate for the context.

Definition: A random variable $X$ has a Poisson distribution with parameter $\lambda>0$, written $X\sim\operatorname{Poisson}(\lambda)$, if for each nonnegative integer $k$, $$ P(X=k)=\frac{e^{-\lambda}\lambda^k}{k!}. $$ Here $k$ is the specified count, $\lambda$ is the expected number of occurrences in the defined window or exposure unit, $e$ is the base of the natural logarithm, and $k!=k(k-1)\cdots 2\cdot1$, with $0!=1$.

The units and context of $\lambda$ matter. If $\lambda=2.4$ calls per ten minutes, the matching random variable counts calls in ten minutes—not in an unspecified period. In a Poisson model, $\lambda$ is the expected count for that window; it is not the probability of an occurrence. It may exceed 1, whereas a probability cannot.

To find the probability of exactly $k$, substitute the parameter for the stated window and the requested count into the formula. The factorial applies to the count $k$, not to $\lambda$. The exponential term is $e^{-\lambda}$, and the resulting probability must be between 0 and 1.

Define the count.
State what $X$ counts and identify the fixed window or exposure unit.

Match $\lambda$ to the window.
Use the expected count for precisely the same window as the question.

Substitute the exact count.
For a question asking for exactly $k$, calculate $e^{-\lambda}\lambda^k/k!$.

Interpret the result.
Report it as the probability of that exact count in context, with sensible rounding.

Worked Calculations

Worked Example: Exactly Three Arrivals

Suppose a hypothetical help desk receives calls according to a Poisson model, with an expected 2.4 calls in each ten-minute interval. Find the probability of exactly three calls in a ten-minute interval.

Let $X$ be the number of calls in one ten-minute interval. The model is $X\sim\operatorname{Poisson}(2.4)$, so $\lambda=2.4$ calls per ten minutes and $k=3$. The stipulated model treats arrivals as independent and the average rate as stable over the interval.

Substitute into the individual probability formula:

P(X=3)=\frac{e^{-2.4}(2.4)^3}{3!}

Calculate the parts: $(2.4)^3=13.824$, $3!=6$, and $e^{-2.4}\approx0.090718$. Therefore,

P(X=3)=0.090718\left(\frac{13.824}{6}\right) =0.090718(2.304) \approx0.2090.

The probability of exactly three calls in one ten-minute interval is about $0.2090$, or $20.90\%$. This is not the probability of three or fewer calls; that would require adding probabilities for multiple counts.

A software check in R is dpois(3, lambda = 2.4). In Python with SciPy, use scipy.stats.poisson.pmf(3, mu=2.4). Both return approximately $0.2090$. The argument for the probability mass function is the exact count, not a cutoff.

Worked Example: Exactly Two Defects

In a hypothetical quality-control model, the number of defects on one panel follows a Poisson distribution with expected count $0.8$ defects per panel. Find the probability that a panel has exactly two defects.

Let $X$ count defects on one panel. Assume the Poisson model is suitable for this fixed unit of inspection; in particular, the defect-generating process is modeled as independent rather than clustered. Use $\lambda=0.8$ defects per panel and $k=2$.

P(X=2)=\frac{e^{-0.8}(0.8)^2}{2!}

Since $(0.8)^2=0.64$, $2!=2$, and $e^{-0.8}\approx0.449329$,

P(X=2)=0.449329\left(\frac{0.64}{2}\right) =0.449329(0.32) \approx0.1438.

The probability that a panel has exactly two defects is about $0.1438$, or $14.38\%$. The parameter $0.8$ is the expected number of defects per panel, not an assertion that a panel has a $0.8$ probability of a defect.

Worked Example: Exactly One Rare Event

Suppose a hypothetical monitoring model assigns an expected $0.15$ rare events per batch. Find the probability of exactly one event in a batch. Assume the batches have the same defined exposure and that independent occurrences at a stable average rate make the Poisson model reasonable.

Let $X$ be the number of events in one batch. Then $\lambda=0.15$ events per batch and $k=1$. The formula gives

P(X=1)=\frac{e^{-0.15}(0.15)^1}{1!}.

Here $1!=1$ and $e^{-0.15}\approx0.860708$, so

P(X=1)=0.860708(0.15) \approx0.1291.

Thus, the probability of exactly one event in a batch is about $0.1291$, or $12.91\%$. For example, R evaluates this as dpois(1, lambda = 0.15). This computation concerns exactly one event; it does not answer the different question of whether at least one occurs.

A Useful Calculation Check

The probability formula also yields a convenient relationship between probabilities for neighboring counts. Let $p_k=P(X=k)$. For any nonnegative integer $k$,

\frac{p_{k+1}}{p_k} = \frac{e^{-\lambda}\lambda^{k+1}/(k+1)!} {e^{-\lambda}\lambda^k/k!} = \frac{\lambda}{k+1}.

The exponential terms cancel, the powers of $\lambda$ leave one factor of $\lambda$, and $(k+1)!=(k+1)k!$. Therefore,

P(X=k+1)=P(X=k)\frac{\lambda}{k+1}.

This is a calculation shortcut and a check, not a different probability model. For example, with $\lambda=2.4$, the probability of exactly two calls is $e^{-2.4}(2.4)^2/2!\approx0.2613$. The relation gives

P(X=3)\approx0.2613\frac{2.4}{3} \approx0.2090,

which agrees with the direct calculation in the arrivals example, apart from rounding. Each probability here still refers to one exact count.

Common Mistakes and Practical Checks

An exact-count calculation is short, but small setup errors can produce a plausible-looking wrong answer. Before reporting a result, check that the model, count, and parameter all refer to the same context.

Using the wrong window for $\lambda$. If the question concerns one ten-minute interval, a parameter stated per hour cannot be inserted as if it were already the ten-minute expected count. Identify the exposure unit before substituting.
Treating $\lambda$ as a probability. A value such as $2.4$ can be a valid expected count even though it cannot be a probability. The Poisson formula uses it as a parameter.
Using the wrong factorial. For exactly three events, the denominator is $3!=6$, not $3$ and not $\lambda!$. For zero events, $0!=1$.
Confusing an exact count with a range. $P(X=2)$ is only the probability of two. A request for two or fewer or at least two needs a sum or complement involving multiple counts.
Rounding too early. Keep extra digits in $e^{-\lambda}$ and intermediate products, then round the final probability. Check that the result lies between 0 and 1.

Software is useful for checking arithmetic, especially when $k!$ or $\lambda^k$ becomes large. It does not choose the model or interpret the result for you. In R, dpois(k, lambda = ...) evaluates the probability for exactly $k$; in Python SciPy, poisson.pmf(k, mu=...) does the same. Use a probability mass function command rather than a cumulative function when the question asks for one specified count.

Key takeaway: For a Poisson count $X$ with parameter $\lambda$, calculate the probability of exactly $k$ occurrences as $e^{-\lambda}\lambda^k/k!$. Define the observation window, match $\lambda$ to it, and state the result as the probability of that exact count in context.

Check Your Understanding

For each question, identify the exact count and the appropriate Poisson parameter before calculating.

A hypothetical service receives an expected 1.6 requests per hour. Write the formula for the probability of exactly two requests in one hour.
For $\lambda=0.5$, calculate $P(X=0)$. What value of $0!$ is used?
A model gives an expected 3 defects per container. Explain why $\lambda=3$ is not a probability.
For a Poisson variable with $\lambda=2$, use the neighboring-count relation to express $P(X=2)$ in terms of $P(X=1)$.
In your own words, distinguish the question answered by $P(X=4)$ from the question “four or fewer.”