Guide

Hypergeometric Distribution Fundamentals

Was this lesson helpful?

Dive Deeper

Generate a follow-up sub-lesson on any aspect of this topic

Guide complete

Hypergeometric Distribution Fundamentals

Build a reliable instinct for sampling without replacement by learning what the hypergeometric distribution counts, how its probability formula is built from combinations, and how mean and variance behave when draws are dependent. Then practice mapping word problems to $N,K,n$ and the event you need.

You grab $n$ items from a box that has a fixed mix of good and defective parts, and you do not put anything back. The number of defectives you end up holding is not binomial, because each draw changes what is left. Hypergeometric distribution fundamentals start with that idea and turn it into a repeatable setup you can use in probability problems.

Let’s make the story concrete before we touch any formulas:

What the hypergeometric distribution models

The hypergeometric distribution models the count of successes in a fixed number of draws when you sample without replacement from a finite population.

You always have the same four roles in the story.

A population of size $N$
$K$ successes in that population
You draw $n$ items without replacement
$X$ is how many successes you drew

The key feature is dependence. After you draw one success, there is one fewer success left, so the chance of success on the next draw changes. That single detail is what makes hypergeometric different from the binomial model.

Same box, shrinking odds
Without replacement means probabilities update after every draw, even if you never compute them step by step.

Recognizing when to use it

Use hypergeometric when the problem sounds like count successes in $n$ draws from a known finite set, and nothing goes back in.

Clues that strongly point to hypergeometric:

The population is explicitly limited, like a deck of 52 or a batch of 200
You draw a fixed number $n$ without replacement
Success is a category label, like red cards, defective items, or students who passed
The question asks for $P(X=x)$ , at least one, at most two, or between

Binomial is the close neighbor. Binomial fits sampling with replacement, or independent trials with a constant success probability $p$ . Hypergeometric fits draws that change the composition of what remains.

Approximations can be fine when the population is large and you sample only a small fraction. If $n$ is small compared to $N$ , dependence becomes weak and the binomial with $p=K/N$ often tracks well.

Here is a compact comparison to lock in the pattern:

The probability mass function in one line

For $X\sim\text{Hypergeometric}(N,K,n)$ , the probability of exactly $x$ successes is

P(X=x)=\frac{\binom{K}{x}\binom{N-K}{n-x}}{\binom{N}{n}}.

This works because each term counts ways to pick sets, not sequences.

$\binom{K}{x}$ counts ways to choose which $x$ successes you take from the $K$ available.
$\binom{N-K}{n-x}$ counts ways to choose the remaining $n-x$ items from the failures.
$\binom{N}{n}$ counts all equally likely sets of $n$ items you could have drawn.

The allowed values of $x$ are only those that make sense physically. You cannot draw more successes than exist, and you cannot draw more successes than draws, so $0\le x\le \min(K,n)$ . Also, if failures are scarce you might be forced to draw some successes, which raises the minimum.

To see how each piece matches the story, walk through the labeled formula:

How the parameters change the shape

Even without graphing, the distribution has a simple center. It clusters around $n\cdot(K/N)$ , since $K/N$ is the overall success fraction in the population.

What changes the look:

Changing $n$

As $n$ increases, $X$ gets more concentrated relative to its range. You are averaging over more draws, so results become more predictable. Also, extreme outcomes become constrained. If you draw almost the whole population, $X$ becomes almost fixed.

Changing the success rate $K/N$

If $K/N$ is near 0.5, the distribution tends to be more symmetric. If successes are rare, the mass stacks near 0 and the right tail thins. If successes are common, the mirror happens near $n$ .

Changing $N$ while keeping $K/N$ fixed

A larger $N$ with the same fraction makes dependence weaker for a given $n$ . The shape starts to resemble a binomial more closely because removing one item barely changes the remaining mix.

A quick simulation makes these shifts obvious:

Sampling fraction matters
When $n/N$ is tiny, without replacement behaves almost like with replacement.

Mean and variance intuition

The mean is exactly what your intuition suggests.

E[X]=n\frac{K}{N}.

Think of $K/N$ as the population success proportion. Over $n$ draws, you expect that proportion of successes.

Variance is where without replacement leaves a fingerprint. For hypergeometric,

\operatorname{Var}(X)=n\frac{K}{N}\left(1-\frac{K}{N}\right)\frac{N-n}{N-1}.

The extra factor $\frac{N-n}{N-1}$ is the finite population correction. It is always at most 1, so variance is smaller than the binomial variance $np(1-p)$ with $p=K/N$ .

Why it shrinks: if you already drew more successes than expected early on, there are fewer successes left, which pulls later draws back toward balance. The population itself acts like a stabilizer.

To uncover where that correction comes from and how to use it, reveal it step by step:

Worked problem pattern you can reuse

Most mistakes happen before any arithmetic. A good setup is half the solution.

Framework labels

Population identifies the full set you are drawing from, and fixes $N$ .
Success names the category you are counting, and fixes $K$ .
Draws is the sample size, and fixes $n$ .
Random variable is $X$ , the number of successes in the sample.
Target event states what probability you want, like $P(X=3)$ or $P(X\ge2)$ .

A common trap is mixing up $K$ and $n$ . $K$ is how many successes exist in the whole population. $n$ is how many items you draw.

Once mapped, compute exactly what the question asks.

For exactly $x$ , use the PMF directly.
For at least or at most, add the appropriate PMF values.
For between, add a short range of $x$ values, not everything.

If the range is large, it can be faster to use a complement such as $P(X\ge2)=1-P(X=0)-P(X=1)$ .

Try converting a word problem into $N,K,n,X$ and a target probability:

Was this lesson helpful?

Dive Deeper

Generate a follow-up sub-lesson on any aspect of this topic

Guide complete

Hypergeometric Distribution Fundamentals

Build a reliable instinct for sampling without replacement by learning what the hypergeometric distribution counts, how its probability formula is built from combinations, and how mean and variance behave when draws are dependent. Then practice mapping word problems to $N,K,n$ and the event you need.

Let’s make the story concrete before we touch any formulas:

What the hypergeometric distribution models

The hypergeometric distribution models the count of successes in a fixed number of draws when you sample without replacement from a finite population.

You always have the same four roles in the story.

A population of size $N$
$K$ successes in that population
You draw $n$ items without replacement
$X$ is how many successes you drew

Same box, shrinking odds
Without replacement means probabilities update after every draw, even if you never compute them step by step.

Recognizing when to use it

Use hypergeometric when the problem sounds like count successes in $n$ draws from a known finite set, and nothing goes back in.

Clues that strongly point to hypergeometric:

The population is explicitly limited, like a deck of 52 or a batch of 200
You draw a fixed number $n$ without replacement
Success is a category label, like red cards, defective items, or students who passed
The question asks for $P(X=x)$ , at least one, at most two, or between

Here is a compact comparison to lock in the pattern:

The probability mass function in one line

For $X\sim\text{Hypergeometric}(N,K,n)$ , the probability of exactly $x$ successes is

P(X=x)=\frac{\binom{K}{x}\binom{N-K}{n-x}}{\binom{N}{n}}.

This works because each term counts ways to pick sets, not sequences.

$\binom{K}{x}$ counts ways to choose which $x$ successes you take from the $K$ available.
$\binom{N-K}{n-x}$ counts ways to choose the remaining $n-x$ items from the failures.
$\binom{N}{n}$ counts all equally likely sets of $n$ items you could have drawn.

To see how each piece matches the story, walk through the labeled formula:

How the parameters change the shape

Even without graphing, the distribution has a simple center. It clusters around $n\cdot(K/N)$ , since $K/N$ is the overall success fraction in the population.

What changes the look:

Changing $n$

Changing the success rate $K/N$

If $K/N$ is near 0.5, the distribution tends to be more symmetric. If successes are rare, the mass stacks near 0 and the right tail thins. If successes are common, the mirror happens near $n$ .

Changing $N$ while keeping $K/N$ fixed

A larger $N$ with the same fraction makes dependence weaker for a given $n$ . The shape starts to resemble a binomial more closely because removing one item barely changes the remaining mix.

A quick simulation makes these shifts obvious:

Sampling fraction matters
When $n/N$ is tiny, without replacement behaves almost like with replacement.

Mean and variance intuition

The mean is exactly what your intuition suggests.

E[X]=n\frac{K}{N}.

Think of $K/N$ as the population success proportion. Over $n$ draws, you expect that proportion of successes.

Variance is where without replacement leaves a fingerprint. For hypergeometric,

\operatorname{Var}(X)=n\frac{K}{N}\left(1-\frac{K}{N}\right)\frac{N-n}{N-1}.

The extra factor $\frac{N-n}{N-1}$ is the finite population correction. It is always at most 1, so variance is smaller than the binomial variance $np(1-p)$ with $p=K/N$ .

Why it shrinks: if you already drew more successes than expected early on, there are fewer successes left, which pulls later draws back toward balance. The population itself acts like a stabilizer.

To uncover where that correction comes from and how to use it, reveal it step by step:

Worked problem pattern you can reuse

Most mistakes happen before any arithmetic. A good setup is half the solution.

Framework labels

Population identifies the full set you are drawing from, and fixes $N$ .
Success names the category you are counting, and fixes $K$ .
Draws is the sample size, and fixes $n$ .
Random variable is $X$ , the number of successes in the sample.
Target event states what probability you want, like $P(X=3)$ or $P(X\ge2)$ .

A common trap is mixing up $K$ and $n$ . $K$ is how many successes exist in the whole population. $n$ is how many items you draw.

Once mapped, compute exactly what the question asks.

For exactly $x$ , use the PMF directly.
For at least or at most, add the appropriate PMF values.
For between, add a short range of $x$ values, not everything.

If the range is large, it can be faster to use a complement such as $P(X\ge2)=1-P(X=0)-P(X=1)$ .

Try converting a word problem into $N,K,n,X$ and a target probability:

Guide complete

Related content

Hypergeometric Distribution Fundamentals

What the hypergeometric distribution models

Recognizing when to use it

The probability mass function in one line

How the parameters change the shape

Changing $n$

Changing the success rate $K/N$

Changing $N$ while keeping $K/N$ fixed

Mean and variance intuition

Worked problem pattern you can reuse

Next steps

Guide complete

Related content

Hypergeometric Distribution Fundamentals

What the hypergeometric distribution models

Recognizing when to use it

The probability mass function in one line

How the parameters change the shape

Changing $n$

Changing the success rate $K/N$

Changing $N$ while keeping $K/N$ fixed

Mean and variance intuition

Worked problem pattern you can reuse

Next steps

Guide complete

Related content

Hypergeometric Distribution Fundamentals

What the hypergeometric distribution models

Recognizing when to use it

The probability mass function in one line

How the parameters change the shape

Changing nnn

Changing the success rate K/NK/NK/N

Changing NNN while keeping K/NK/NK/N fixed

Mean and variance intuition

Worked problem pattern you can reuse

Next steps

Guide complete

Related content

Hypergeometric Distribution Fundamentals

What the hypergeometric distribution models

Recognizing when to use it

The probability mass function in one line

How the parameters change the shape

Changing nnn

Changing the success rate K/NK/NK/N

Changing NNN while keeping K/NK/NK/N fixed

Mean and variance intuition

Worked problem pattern you can reuse

Next steps

Changing $n$

Changing the success rate $K/N$

Changing $N$ while keeping $K/N$ fixed

Changing $n$

Changing the success rate $K/N$

Changing $N$ while keeping $K/N$ fixed