What is the Hypergeometric Distribution?

The hypergeometric distribution describes outcomes when sampling without replacement. Suppose your population contains a fixed number of items, some with a desired property (successes) and others without. When you draw a sample, the probability of success changes after each item is removed—unlike the binomial distribution where replacement keeps odds constant.

This distribution appears whenever items cannot be redrawn: selecting employees for a committee, inspecting parts from a batch, drawing cards from a deck, or evaluating defective units in manufacturing. The key constraint is sampling without replacement, making hypergeometric probability essential for finite-population scenarios.

The Hypergeometric Probability Formula

Three parameters fully define the hypergeometric distribution:

  • N — total population size
  • K — number of success states in the population
  • n — sample size drawn
  • k — number of successes observed in the sample

The probability mass function for exactly k successes is:

P(X = k) = C(K,k) × C(N−K, n−k) / C(N,n)

where C(a,b) = a! / (b!(a−b)!)

  • N — Total number of items in the population
  • K — Number of items with the desired property in the population
  • n — Number of items drawn in a single sample
  • k — Number of successes (desired property items) in the drawn sample

Mean, Variance, and Distribution Properties

Two key descriptive statistics characterize any hypergeometric distribution:

Mean (Expected Value): The average number of successes in a sample is proportional to the success rate in the population.

μ = n × K / N

Variance: Variability depends on both the success proportion and the ratio of sample size to population size. Smaller populations or larger samples relative to the population reduce variance.

σ² = n × (K/N) × ((N−K)/N) × ((N−n)/(N−1))

Notice the correction factor (N−n)/(N−1) that accounts for sampling without replacement. As N grows very large, this factor approaches 1, and hypergeometric probabilities converge toward binomial probabilities.

Hypergeometric vs. Binomial Distribution

Both distributions calculate success probabilities over multiple trials, but differ fundamentally:

  • Hypergeometric: Sampling without replacement. Each draw removes an item, shifting the population composition and success probability for the next draw.
  • Binomial: Sampling with replacement (or infinite population). Success probability remains constant across all trials.

In practice, when sample size is much smaller than population size (rule of thumb: n < 5% of N), hypergeometric and binomial results are nearly identical, and binomial calculations become easier. For small populations or large samples relative to population, hypergeometric is the only correct choice.

Common Pitfalls When Using Hypergeometric Probabilities

Avoid these frequent mistakes when applying hypergeometric calculations:

  1. Confusing sampling with and without replacement — Hypergeometric assumes you cannot redraw the same item. If you're sampling with replacement, use binomial distribution instead. This distinction matters most when the population is small or the sample is large relative to N.
  2. Incorrectly counting success states in the population — Ensure K (successes in population) does not exceed N, and that K ≥ 0. Similarly, verify that n ≤ N and that the maximum possible successes in the sample cannot exceed min(K, n). Miscounting population successes invalidates all downstream calculations.
  3. Applying hypergeometric to large populations unnecessarily — When N is very large (thousands or millions) and n is relatively small, computational complexity increases without meaningful accuracy gain over binomial. Use binomial as a practical approximation in these scenarios, or limit precision expectations due to rounding in factorials.
  4. Misinterpreting cumulative probabilities — P(X ≥ k) means k or more successes. P(X ≤ k) means k or fewer. Ensure you select the correct cumulative direction. The tail direction matters for hypothesis testing and decision-making.

Frequently Asked Questions

When should I use hypergeometric distribution instead of binomial?

Use hypergeometric whenever you sample without replacement from a finite population. Classic examples include quality control inspections (testing items from a batch without putting them back), lottery drawings, card games, and hiring committees. Use binomial only when the population is infinite or when you sample with replacement. A practical rule: if removing one item noticeably changes the probability of success for the next draw, hypergeometric is correct.

How does the hypergeometric mean relate to the binomial mean?

The hypergeometric mean is μ = n × K/N, where K/N is the population success rate. The binomial mean is μ = n × p, where p is the constant success probability. If you treat K/N as the population proportion p in a binomial, the means are identical. However, hypergeometric variance is smaller because removing items reduces variability—the correction factor (N−n)/(N−1) makes this explicit.

What's the difference between P(X=k) and P(X≤k) in this calculator?

P(X=k) is the exact probability of observing exactly k successes. P(X≤k) is the cumulative probability of getting k or fewer successes (includes 0, 1, 2, ..., k). Use P(X=k) for specific outcomes and P(X≤k) for threshold decisions. Similarly, P(X≥k) gives k or more successes, useful for "at least" questions common in quality and risk assessment.

Can hypergeometric probabilities exceed the binomial approximation?

No. Hypergeometric and binomial probabilities are always between 0 and 1. However, hypergeometric variance is always less than or equal to the binomial variance for identical parameters, due to the correction factor. When the population is finite and small relative to the sample, this difference becomes pronounced. For very large populations (N > 10,000) relative to sample size, the distributions are nearly indistinguishable.

What happens if K is very small compared to N?

If K ≪ N (few successes in a large population), hypergeometric probabilities approach Poisson distribution behavior. The mean μ = n × K/N remains low, and the probability of observing any success becomes small. Variance still follows the hypergeometric formula but is suppressed by the (N−K)/N term, which approaches 1. This scenario is common in rare-event detection and defect-rate studies.

How do I interpret variance in the context of sampling without replacement?

Variance measures spread around the mean number of successes. Higher variance means the actual number of successes could vary widely; lower variance suggests more consistent outcomes. Sampling without replacement reduces variance because late draws are constrained by earlier removals—you cannot draw the same success twice. This makes outcomes more predictable than with-replacement binomial sampling, an advantage in batch testing and inventory control.

More statistics calculators (see all)