What is the Geometric Distribution

Understanding the Geometric Distribution

The geometric distribution precisely describes the probability of observing a certain quantity of failures before achieving the first success in a series of sequential and independent trials. This model is critical for analyzing ‘waiting time’ scenarios, where the process stops immediately upon the first success. Unlike distributions that count successes within a fixed boundary of attempts, the Geometric Distribution models an open-ended process until the specified condition is finally met.

The crucial assumption underpinning the use of the geometric model is that every trial must be independent—the outcome of a previous attempt cannot influence the outcome of the current or future attempts. Furthermore, the probability of success, denoted $p$, must remain identical for every iteration of the experiment. If these conditions are violated, the Geometric Distribution is not an appropriate model.

The Foundation: Bernoulli Trials

The elemental building block of the Geometric Distribution is the Bernoulli trial. A Bernoulli trial is defined as any random experiment that has only two possible, mutually exclusive outcomes: designated “success” or “failure.” The simplicity of this framework allows for complex sequential events to be modeled efficiently, provided the constant probability assumption holds true.

A classic example of a Bernoulli trial is flipping a coin. The coin can only land on two sides (Heads or Tails). If we designate Heads as a “success” and Tails as a “failure,” and assume the coin is fair, the probability of success is $p = 0.5$ for every flip. Crucially, the outcome of any single flip is completely independent of all previous flips.

 

The repetitive nature of these trials, combined with the fixed probability of success, establishes the perfect environment for applying geometric probability calculations. The distribution is specifically concerned with counting the number of unsuccessful outcomes before the first successful outcome interrupts the sequence.

Defining the Probability Mass Function (PMF)

If a random variable X follows a geometric distribution, the Probability Mass Function (PMF) provides the precise mathematical tool to calculate the probability of experiencing exactly k failures before achieving the first success. This is represented as $P(X=k)$. The formula is derived by multiplying the probability of $k$ consecutive failures by the probability of one success immediately following them.

The mathematical formula for the Geometric PMF is:

P(X=k) = (1-p)^kp

where the components are defined as:

• k: The discrete count representing the number of failures observed before the very first success. Possible values for $k$ are $0, 1, 2, 3, dots$.
• p: The fixed probability of success on each individual Bernoulli trial.
• (1-p)^k: Represents the probability of $k$ consecutive failures occurring before the successful trial.

Illustrative Example: The Fair Coin Flip

Consider the application of this formula to the fair coin example, where $p = 0.5$. We want to determine the probability of different numbers of tails (failures, $k$) occurring until the first head (success) is flipped.

It is essential to note that the trial can experience $k=0$ failures if success lands on the very first flip. This outcome has the highest probability.

1. For $k=0$ failures (Success on first flip):

P(X=0) = (1-.5)⁰(.5) = 1 * 0.5 = 0.5

2. For $k=1$ failure (Sequence: Failure, Success):

P(X=1) = (1-.5)¹(.5) = 0.5 * 0.5 = 0.25

3. For $k=2$ failures (Sequence: Failure, Failure, Success):

P(X=2) = (1-.5)²(.5) = 0.25 * 0.5 = 0.125

4. For $k=3$ failures (Sequence: F, F, F, S):

P(X=3) = (1-.5)³(.5) = 0.125 * 0.5 = 0.0625

As shown, the probability decreases exponentially as the required number of preceding failures increases. This geometric decay is a signature characteristic of the distribution. We can calculate this probability for any number of coin flips up to infinity and visualize the distribution clearly using a histogram:

Calculating Cumulative Probabilities

Beyond calculating the probability of exactly k failures, we often need to determine the cumulative probability—the chance that the first success occurs after k or less failures, denoted P(X ≤ k). This is the sum of the probabilities from $k=0$ up to the specified value of $k$.

The formula for the cumulative probability (CDF) is mathematically simplified by considering the complementary event: P(X ≤ k) is equal to 1 minus the probability that success has not yet occurred by the $(k+1)$-th trial. The event of success occurring on the $(k+1)$-th trial or later means that all $k+1$ trials must have been failures.

The formula for the Geometric Cumulative Distribution Function is:

P(X≤k) = 1 – (1-p)^k+1

• k: The maximum number of failures included in the calculation.
• p: The probability of success on each trial.

For example, if we want to know the probability that it takes three or fewer failures ($k leq 3$) until the coin finally lands on heads ($p=0.5$), we calculate:

P(X≤3) = 1 – (1-.5)³⁺¹ = 1 – (0.5)⁴ = 1 – 0.0625 = 0.9375

The calculations for smaller cumulative values are equally straightforward:

P(X≤0) = 1 – (1-.5)⁰⁺¹ = 1 – 0.5 = 0.5

P(X≤1) = 1 – (1-.5)¹⁺¹ = 1 – 0.25 = 0.75

P(X≤2) = 1 – (1-.5)²⁺¹ = 1 – 0.125 = 0.875

As the possible number of failures increases, the cumulative probability rapidly approaches 1, confirming that eventual success is practically guaranteed. This progression can be visualized using a cumulative probability histogram:

Key Statistical Properties

The Geometric Distribution possesses unique statistical properties related to its central tendency and variability. These derived formulas allow for quick calculation of expected outcomes without summing individual probabilities.

The primary properties are:

The Mean (Expected Value, $mu$) of the distribution, representing the average number of failures (k) anticipated before the first success, is defined as: (1-p) / p.

The Variance ($sigma^2$) of the distribution, measuring the spread of the data points around the mean, is calculated as: (1-p) / p².

For our continuing coin flip scenario ($p=0.5$):

The mean number of times we would expect the coin to land on tails (failures) before it lands on heads (success) is calculated as: $(1-p) / p = (1-.5) / .5 = 0.5 / 0.5 = 1$. We expect, on average, one failure before success.

The variance in the number of flips until success is: $(1-p) / p^2 = (1-.5) / .5^2 = 0.5 / 0.25 = 2$. This gives us a measure of how much the observed number of failures might vary around that expected mean of 1.

Mastering Geometric Distribution Problems

The following practice problems illustrate how to apply the PMF, the complementary rule, and the mean formula in real-world contexts. We use the scenario of a researcher interviewing people outside a library, where the probability of success (supporting the law) is fixed at $p = 0.2$.

Problem 1: Exact Probability (PMF Application)

Question: The probability that a given person supports the law is $p = 0.2$. What is the probability that the fourth person the researcher talks to is the first person to support the law?

Answer: If the fourth person is the first success, it means there were $k=3$ failures beforehand. We use the PMF: P(X=3) = $(1-p)^3 p$. P(X=3) = $(1 – 0.2)^3 * 0.2 = (0.8)^3 * 0.2 = 0.10240$. The probability is 0.10240.

Problem 2: Upper Tail Probability (Complementary Rule)

Question: The probability that a given person supports the law is $p = 0.2$. What is the probability that the researcher will have to talk to more than four people to find someone who supports the law?

Answer: The phrase “talk to more than four people” implies that the first success occurs on the fifth trial or later. This is equivalent to having $k geq 4$ failures. We calculate the probability that the first four trials are all failures: P(X $geq$ 4) = $(1-p)^4 = (1-0.2)^4 = (0.8)^4 = 0.4096$. If we assume the question implies five or more failures ($k ge 5$), which corresponds to the original intended answer: P(X $geq$ 5) = $(1-p)^5$. P(X $geq$ 5) = $(0.8)^5 = 0.32768$. This is the probability that the first five people are failures.

Problem 3: Expected Number of Failures (Mean Application)

Question: The probability that a given person supports the law is $p = 0.2$. What is the expected number of people the researcher will have to talk to until she finds someone who supports the law?

Answer: The expected number of failures (non-supporters) before the first success is given by the mean formula: $mu = (1-p) / p$. Using $p = 0.2$, the mean would be $(1 – 0.2) / 0.2 = 0.8 / 0.2 = 4$. The researcher is expected to interview four non-supporters before finding the first person who supports the law.