A sampling distribution is the probability distribution of a statistic — usually the sample mean — computed over all possible random samples of a fixed size from a population.
σxˉ=nσ
Quick Answer
A sampling distribution is the probability distribution of a statistic — usually the sample mean — computed over all possible random samples of a fixed size from a population. Use it when you ask how much a sample average, or a sample proportion, would change if you drew a different random sample. Its center is the population value, and its spread, the standard error (nσ for a mean), shrinks as samples get bigger. Every margin of error and every p-value is built on it.
Why This Matters
A poll of 1,000 people gives one percentage; a second poll would give a slightly different one. Without a sampling distribution you can't tell whether a 3-point change is news or noise. With it you know how far sample results typically wander from the truth, which is the basis of confidence intervals, margins of error and hypothesis tests. Confusing it with the spread of the raw data makes estimates look far less precise than they are.
Intuitive Explanation
Suppose adult heights in a city have mean μ=170 cm and standard deviationσ=10 cm. Pick 25 people at random and average their heights: maybe 171.8. Pick another 25: maybe 168.9. Repeat thousands of times and draw a histogram of the averages, not of the people. That histogram is the sampling distribution of xˉ.
Two things stand out. The averages center on 170, the population mean, because random samples are not tilted up or down. And they cluster far more tightly than individual heights: one person can easily be 15 cm from 170, but an average of 25 people rarely lands more than 4 cm away, because tall and short people in the same sample offset each other. The spread of the averages is the standard error, nσ=2510=2 cm.
Bigger samples shrink this spread, but slowly: halving the standard error takes four times the sample, since 4=2. Notice what does not shrink. Individual heights still vary with σ=10 cm however many people you measure; the population distribution stays put. Sample size changes how much averages wobble, not how much people differ.
Think of it like this.Imagine many reporters, each polling a different random group of 25 people and announcing the group's average. The sampling distribution is the spread of their headlines. Individual people disagree a lot; the reporters' averages disagree much less, and the more people each reporter asks, the closer the headlines crowd around the truth.
Core idea.A sampling distribution is the spread of a statistic (like the sample mean) over all possible samples of the same size.
When to Use
Use the sampling distribution when a question is about a statistic, such as a sample mean or sample proportion, and how it would vary across repeated samples: how far an average of 40 might fall from the true mean, or how likely a sample mean above some value is. Use σ for a question about one individual and nσ for a question about an average of n. To justify a normal shape when the population is skewed, add the central limit theorem.
Tip.Ask: Am I describing how a statistic (like xˉ) varies across many samples, rather than how raw values vary?
How to Recognize It
Before choosing a standard deviation, decide whether the question is about one individual, one sample you already have, or an average that would change if you sampled again.
1.
Is the question about the average of a group of n, such as the mean weight of 25 apples, rather than one apple?
Yes means use the sampling distribution, with standard error nσ. A question about a single individual uses the population distribution and σ itself.
2.
Does the question imagine drawing the sample again and ask how much the result could change?
Yes means you need the spread of the statistic across repeated samples. If it asks only to describe the data you already collected, such as their mean or range, you are working with a single sample.
3.
Does the question want the shape of that distribution, in particular whether it is close to normal?
Yes brings in the central limit theorem: with a large sample, the sample mean is close to normal even when the population is skewed. If the population is normal to begin with, the sample mean is normal for every n.
4.
Is the statistic a proportion, like the share of 400 voters who back a candidate, rather than a mean?
Yes means the same idea with a different spread: the sample proportion p^ is centered at p with standard error np(1−p), which also shrinks as n grows.
Sampling Distribution compared with Population distribution, Central limit theorem, and Sample (single)
All four rows describe a spread or shape, but of different things. Ask what each dot would be: one individual (population distribution), one value you collected (a single sample), or the mean of one of many possible samples (sampling distribution, whose normal shape the CLT explains).
Sampling Distribution compared with Population distribution, Central limit theorem, and Sample (single)
Type
Meaning
Key test
Formula
Example
Sampling Distribution
Use this when you need the distribution of a sample statistic across repeated samples, not the distribution of individual values.
Am I describing how a statistic (like xˉ) varies across many samples, rather than how raw values vary?
σxˉ=nσ
A population has mean μ=170 cm and SD σ=10 cm. For samples of size n=25, what is the standard deviation of the sample mean?
Population distribution
The spread of individual values in the whole population.
Use when describing the data itself, not a statistic over samples.
σ
All adult heights
Central limit theorem
The result that the sampling distribution of the mean is ~normal for large n.
Use when invoking WHY the sampling distribution is bell-shaped, not defining it.
Xˉ∼N(μ,nσ)
Means of large samples look normal
Sample (single)
One drawn set of data; the sampling distribution is over ALL such samples.
Use when working with the data you actually collected, not the hypothetical many.
—
The one survey of 50 you ran
Sampling Distribution
Meaning
Use this when you need the distribution of a sample statistic across repeated samples, not the distribution of individual values.
Key test
Am I describing how a statistic (like xˉ) varies across many samples, rather than how raw values vary?
Formula
σxˉ=nσ
Example
A population has mean μ=170 cm and SD σ=10 cm. For samples of size n=25, what is the standard deviation of the sample mean?
Population distribution
Meaning
The spread of individual values in the whole population.
Key test
Use when describing the data itself, not a statistic over samples.
Formula
σ
Example
All adult heights
Central limit theorem
Meaning
The result that the sampling distribution of the mean is ~normal for large n.
Key test
Use when invoking WHY the sampling distribution is bell-shaped, not defining it.
Formula
Xˉ∼N(μ,nσ)
Example
Means of large samples look normal
Sample (single)
Meaning
One drawn set of data; the sampling distribution is over ALL such samples.
Key test
Use when working with the data you actually collected, not the hypothetical many.
Example
The one survey of 50 you ran
Formula & Notation
σxˉ=nσ
E(Xˉ)=μ and SD(Xˉ)=nσ where μ and σ are the population mean and SD
How to read it:Xˉ denotes the random variable for the sample mean; its distribution is the sampling distribution.
Worked Examples
Example 1 — Spread of the sample mean
Easy
A population has mean μ=170 cm and SD σ=10 cm. For samples of size n=25, what is the standard deviation of the sample mean?
Show the steps and the answer
Note that the question is about the sample mean, not one person.
So the spread needed is the standard error nσ, not σ itself.
Take the square root of the sample size: 25=5.
Averaging 25 independent values cuts the spread by a factor of 25.
Divide: σxˉ=510=2 cm.
Sample means from groups of 25 typically land within about 2 cm of 170.
Answer:Standard error =2 cm
Takeaway:The standard error nσ measures how much averages wobble; here it is a fifth of the spread of individuals.
Example 2 — Spread of the data, not the mean
Standard
In the same population, how much do individual heights vary?
Show the steps and the answer
Notice that the question is about individual heights, not averages.
Individuals follow the population distribution, so no sampling distribution is involved.
Read the spread straight from the population: σ=10 cm.
There is no division by n, because looking at one person involves no averaging.
Set it beside the previous answer: 10 cm for individuals versus 2 cm for means of 25.
Mixing up these two spreads is the most common error with sampling distributions.
Answer:σ=10 cm
Takeaway:Individuals vary by σ; averages of n vary by nσ. Read whether the question is about one value or an average.
Example 3 — Find the mistake: using sigma for an average
Application
Bags of chips have weights with mean μ=200 g and σ=6 g. An inspector weighs a random sample of 36 bags. Find the probability that the sample mean is at least 203 g. Dev writes: "z=6203−200=0.5, so the probability is about 0.31." Find the mistake and correct it.
Show the steps and the answer
Identify the random quantity: the mean of 36 bags, not one bag.
Dev's z uses σ=6 g, which describes single bags.
Compute the standard error: σxˉ=366=66=1 g.
Averages of 36 bags vary six times less than single bags do.
Recompute z=1203−200=3.
203 g is 3 standard errors above the mean, an extreme value for an average.
Find P(Z≥3)≈0.0013.
With n=36 the sample mean is close to normal, so the normal table applies. Dev's 0.31 is roughly the chance that one bag weighs at least 203 g.
Answer:P(xˉ≥203)=P(Z≥3)≈0.0013, not 0.31
Takeaway:For a question about an average, standardize with nσ; using σ answers a question about a single bag instead.
Where You'll See It
Why two honest polls disagree
If exactly 50% of voters favor a candidate, polls of 1,000 random voters still won't all say 50%. Their sampling distribution has standard error 10000.5×0.5≈0.016, so about 95% of such polls land between roughly 47% and 53%. A one-point swing between two polls sits well within that wobble.
Hourly checks on cereal boxes
Single cereal boxes might vary by σ=8 g, but a quality team averages 16 boxes every hour. Those averages vary by only 168=2 g, so an hourly average 6 g below target, 3 standard errors, is strong evidence that the filling machine has drifted.
Common Mistakes
The slip:Using the population SD σ as the spread of the mean
Instead:the sample mean's spread is the smaller standard error nσ.
The slip:Confusing the data's distribution with the statistic's distribution
Instead:raw values and sample means have different spreads.
The slip:Thinking a bigger sample makes individual data less variable
Instead:bigger n shrinks the spread of the MEAN, not of the data.
Mini Practice
Try these on your own. Tap Reveal when you want to check.
1.
A population has σ=12. What is the standard deviation of xˉ for samples of size n=36?
Hint:Divide σ by the square root of the sample size.
2.
A population has σ=20. Find the standard error of xˉ for n=16 and for n=64. What happens when n is multiplied by 4?
Hint:Compute nσ for each size.
3.
In a population, 40% prefer brand X. For random samples of 100, what is the standard deviation of the sample proportion p^?
Hint:Use np(1−p).
4.
Heights are normal with μ=170 cm and σ=10 cm. Which is more likely: one person taller than 180 cm, or the mean of 25 people above 180 cm?
Hint:Find a z-score for each, using σ for one person and nσ for the mean.
5.
Question A: "How much do individual apple weights vary?" Question B: "How much would the average weight of 50 apples vary from bag to bag?" Which is about a sampling distribution?
Hint:One is about raw values; the other is about a statistic.
6.
A researcher increases n from 25 to 100 and says: "Now the people's heights vary half as much." Correct her.
Hint:Does sampling more people change how different people are?
Want the full set?
50 practice questions for this concept — free to try, every one with a complete worked
solution showing the why, not just the answer.
Why does the spread shrink with n and not with n?
Because random errors partly cancel rather than add up. When you average n values, the highs and lows offset each other, but not perfectly, and the leftover error falls like n1. So to halve the standard error you need four times as many observations: going from 25 to 100 cuts nσ from 5σ to 10σ. Each extra digit of precision costs a hundred times the data.
Does the size of the population matter, or only the size of the sample?
Mostly only the sample size. A random sample of 1,000 gives roughly the same precision whether it comes from a town of 50,000 or a country of 300 million, since nσ has no population size in it. Population size starts to matter only when the sample is a large share of it, more than about 10%; then the true standard error is smaller than the formula says.
If no one ever takes all possible samples, how can we know a sampling distribution?
In three ways. Theory gives it directly: the formulas for the center and the standard error come from probability rules, with no repeated sampling needed. A computer simulation can draw thousands of samples from a model population and plot their means. And a method called the bootstrap resamples, with replacement, from the one sample you actually have, to estimate how much the statistic would vary.
Is the center of a sampling distribution always the population value?
For the sample mean and the sample proportion, yes: averaged over all samples, xˉ equals μ and p^ equals p, which is why they are called unbiased. Not every statistic behaves this way. The sample range, for example, is on average smaller than the population range, because a sample rarely contains both the very largest and the very smallest values.
You need the Normal Distribution, the Mean and Standard Deviation first. The sampling distribution moves attention from individual values to how a statistic behaves across samples, with center μ and spread nσ. Next, the Central Limit Theorem explains its normal shape, Confidence Intervals turn its spread into a range for an unknown mean, and Hypothesis Testing asks whether an observed statistic is unusual under it.