Sample distribution vs sampling distribution

By Jude Wallis · Published

A sample distribution is the distribution of the individual data values in one sample you collected. A sampling distribution is the distribution of a statistic, such as the sample mean, across all possible samples of the same size. One describes data. The other describes how a statistic varies.

Sampling distributions are introduced in Unit 2 topic 2.12, Sampling Distributions and the Central Limit Theorem, in the Fall 2026 AP Statistics course, and the sampling distribution of a sample mean is developed in Unit 4 topic 4.1. Describing the distribution of the data in a sample uses the Unit 1 skills of topics 1.6 and 1.7.

Sample distribution vs sampling distribution: the short answer

The names differ by three letters and the ideas differ in what each dot on the graph stands for.

  • In a sample distribution, also called the distribution of a sample or the distribution of sample data, each value is one individual's measurement. You have it in hand as soon as you collect a sample.
  • In a sampling distribution, each value is a statistic computed from an entire sample: one xˉ\bar{x} ("x-bar"), or one p^\hat{p} ("p-hat"). It describes every sample of that size you could have drawn, so it is a theoretical distribution that you describe with formulas or approximate by simulation.

A sample of 40 students gives a sample distribution made of 40 heights. It contributes exactly one value, its mean, to the sampling distribution of xˉ\bar{x}.

What a sample distribution is

A sample distribution is the set of values of a variable observed in one sample, together with how often each occurs. A dotplot or histogram of your data is a picture of it.

You describe it the way you describe any set of data: shape, center, variability, and unusual features. Its center is a statistic such as the sample mean xˉ\bar{x}, and its variability is a statistic such as the sample standard deviation ss.

When the sample is random, the sample distribution tends to resemble the population it came from. If the population is skewed right, a large random sample from it will usually look skewed right as well, and ss will be close to the population standard deviation σ\sigma. Taking a larger sample gives a more detailed picture of the population. It does not make the data less spread out.

What a sampling distribution is

A sampling distribution is the distribution of the values of a statistic for all possible samples of a given size from a given population. Fix the sample size nn, consider every sample that could be drawn, compute the statistic for each, and collect the results.

For the sample mean xˉ\bar{x} from a population with mean μ\mu and standard deviation σ\sigma, with independent observations:

  • Center: μxˉ=μ\mu_{\bar{x}} = \mu.
  • Variability: σxˉ=σn\sigma_{\bar{x}} = \frac{\sigma}{\sqrt{n}}, which shrinks as nn grows.
  • Shape: normal if the population is normal, and approximately normal for large nn whatever the population's shape, by the central limit theorem.

You almost never observe a sampling distribution directly, because you take one sample, not all of them. You can approximate one by simulation, as the sampling distribution simulator does. A fuller introduction is in sampling distributions explained.

The differences side by side

FeatureSample distributionSampling distribution
Each value isOne individual's data valueOne statistic from a whole sample
Built fromOne sampleAll possible samples of size nn
Observed or theoreticalObserved dataTheoretical, or approximated by simulation
Centerxˉ\bar{x}, which estimates μ\muμ\mu exactly, for the sample mean
Variabilityss, which estimates σ\sigmaσn\frac{\sigma}{\sqrt{n}}, for the sample mean
ShapeTends to resemble the populationApproximately normal for large nn, for the sample mean
Effect of a larger nnResembles the population more closelyBecomes narrower
Used forDescribing the data, checking conditionsComputing probabilities for a statistic, inference

A third distribution sits behind both: the population distribution, which holds the values of the variable for every individual in the population. The sample distribution is a partial copy of it. The sampling distribution is derived from it. The comparison with the model for a single observation is in probability distribution vs sampling distribution.

When to use which

Use the sample distribution when you are describing or checking the data you collected. Reporting the mean and standard deviation of your sample, drawing a boxplot, and looking for strong skew or outliers before a tt procedure are all statements about the sample distribution.

Use the sampling distribution when you are asking how a statistic behaves. "What is the probability that the mean of 36 wait times exceeds 10 minutes" is a question about the sampling distribution of xˉ\bar{x}, and it uses σn\frac{\sigma}{\sqrt{n}}. Standard errors, confidence intervals, and p-values all rest on a sampling distribution.

The two work together in inference. You inspect the sample distribution to judge whether a normal model for the sampling distribution is reasonable, and then you use the sampling distribution to do the calculation.

The classic mix-up and how to avoid it

The first error is saying the sample becomes normal when nn is large. The central limit theorem is a statement about the sampling distribution of the sample mean. A large random sample from a skewed population is still skewed, because it resembles the population. The distribution that becomes approximately normal is the distribution of xˉ\bar{x} across samples.

The second error is saying a larger sample has less variability. The spread of the data, measured by ss, stays near σ\sigma however large the sample is. The quantity that shrinks is the variability of the statistic, σn\frac{\sigma}{\sqrt{n}}. Name which distribution you mean: "the sampling distribution of the sample mean has less variability when nn is larger."

The third error is using the wrong standard deviation in a z-score. A question about one individual uses σ\sigma. A question about a sample mean uses σn\frac{\sigma}{\sqrt{n}}. That choice is the subject of standard error vs standard deviation.

A quick test settles which distribution a graph shows: ask what one dot represents. If a dot is a person, a battery, or a wait time, it is a sample or population distribution. If a dot is an average or a proportion from a whole sample, it is a sampling distribution.

Every possible sample from a tiny population

A spinner lands on 2, 4, 6, or 8 with equal probability. Two spins make a sample of size 2. One particular sample is the pair (2, 8). Describe that sample distribution, then build the sampling distribution of the sample mean and find its mean and standard deviation.

  1. Describe the population: values 2, 4, 6, 8, each with probability 14\frac{1}{4}. Its mean is μ=2+4+6+84=5\mu = \frac{2 + 4 + 6 + 8}{4} = 5.

  2. Population variance: the squared deviations are 9,1,1,99, 1, 1, 9, so σ2=9+1+1+94=5\sigma^2 = \frac{9 + 1 + 1 + 9}{4} = 5 and σ=5≈2.236\sigma = \sqrt{5} \approx 2.236.

  3. The sample distribution for the sample (2, 8) is just those two data values, 2 and 8. Its mean is xˉ=2+82=5\bar{x} = \frac{2 + 8}{2} = 5.

  4. List all possible samples: there are 4×4=164 \times 4 = 16 equally likely ordered pairs. Their means range from 2+22=2\frac{2 + 2}{2} = 2 to 8+82=8\frac{8 + 8}{2} = 8.

  5. Count the pairs for each mean: a mean of 2 comes from 1 pair, 3 from 2 pairs, 4 from 3 pairs, 5 from 4 pairs, 6 from 3 pairs, 7 from 2 pairs, and 8 from 1 pair. Check: 1+2+3+4+3+2+1=161 + 2 + 3 + 4 + 3 + 2 + 1 = 16.

  6. The sampling distribution of xˉ\bar{x} is therefore P(2)=116P(2) = \frac{1}{16}, P(3)=216P(3) = \frac{2}{16}, P(4)=316P(4) = \frac{3}{16}, P(5)=416P(5) = \frac{4}{16}, P(6)=316P(6) = \frac{3}{16}, P(7)=216P(7) = \frac{2}{16}, P(8)=116P(8) = \frac{1}{16}.

  7. Its mean is 5 by symmetry, equal to μ\mu. Its variance is 9(1)+4(2)+1(3)+0(4)+1(3)+4(2)+9(1)16=4016=2.5\frac{9(1) + 4(2) + 1(3) + 0(4) + 1(3) + 4(2) + 9(1)}{16} = \frac{40}{16} = 2.5, so its standard deviation is 2.5≈1.581\sqrt{2.5} \approx 1.581.

  8. Check against the formula: σn=2.2362≈1.581\frac{\sigma}{\sqrt{n}} = \frac{2.236}{\sqrt{2}} \approx 1.581. They agree.

The sample distribution of (2, 8) is two data values with mean 5. The sampling distribution of xˉ\bar{x} for n=2n = 2 has seven possible values, 2 through 8, with probabilities 116,216,316,416,316,216,116\frac{1}{16}, \frac{2}{16}, \frac{3}{16}, \frac{4}{16}, \frac{3}{16}, \frac{2}{16}, \frac{1}{16}. Its mean is 5 and its standard deviation is about 1.581, smaller than the population's 2.236.

A skewed sample and a nearly normal sampling distribution

Wait times at a clinic are strongly skewed right with mean 8 minutes and standard deviation 6 minutes. A random sample of 36 wait times is taken from the clinic's very large set of records. Describe what the sample distribution should look like, describe the sampling distribution of the sample mean, and find the approximate probability that the sample mean exceeds 10 minutes.

  1. Sample distribution: the 36 wait times are a random sample from a right-skewed population, so a histogram of them should also look skewed right, with a mean near 8 minutes and a standard deviation near 6 minutes.

  2. Sampling distribution, center: μxˉ=μ=8\mu_{\bar{x}} = \mu = 8 minutes.

  3. Sampling distribution, variability: the sample is a small fraction of a very large population, so σxˉ=σn=636=66=1\sigma_{\bar{x}} = \frac{\sigma}{\sqrt{n}} = \frac{6}{\sqrt{36}} = \frac{6}{6} = 1 minute.

  4. Sampling distribution, shape: n=36n = 36 is at least 30, so by the central limit theorem the distribution of xˉ\bar{x} is approximately normal even though the population is skewed.

  5. Standardize the sample mean: z=10−81=2z = \frac{10 - 8}{1} = 2.

  6. The area to the right of z=2z = 2 is 1−0.9772=0.02281 - 0.9772 = 0.0228.

  7. Note what cannot be computed this way. The probability that one individual wait exceeds 10 minutes cannot be found from a normal curve, because individual wait times are skewed.

The sample distribution should be skewed right with mean near 8 and standard deviation near 6 minutes. The sampling distribution of xˉ\bar{x} is approximately normal with mean 8 minutes and standard deviation 1 minute, so P(xˉ>10)≈0.0228P(\bar{x} > 10) \approx 0.0228.

Frequently asked questions

Is a sampling distribution made of sample data?

No. A sampling distribution is made of values of a statistic, one value for each possible sample of a given size. The data values inside any one sample make up a sample distribution.

Does a large sample have a normal distribution?

Not necessarily. A large random sample tends to look like its population, so a sample from a skewed population is skewed. The central limit theorem says the sampling distribution of the sample mean is approximately normal for large samples.

Which standard deviation goes with each distribution?

The sample distribution has the sample standard deviation ss, which estimates the population value σ\sigma. The sampling distribution of the sample mean has standard deviation σn\frac{\sigma}{\sqrt{n}}, which is smaller than σ\sigma whenever nn is greater than 1.