A sampling distribution describes how a statistic varies across repeated random samples of the same size. It is different from the distribution of individual observations in one sample and different from the distribution of values in the population.
Keep three distributions separate
The population distribution describes individual values for everyone in the target population. A sample distribution describes the observed individual values in one selected sample. The sampling distribution describes a statistic, such as the sample mean, computed from many hypothetical samples.
For example, individual adult heights vary widely. The mean height of 100 randomly selected adults varies much less from sample to sample. Averaging reduces random variability, even though the underlying individual heights have not become less variable.
A parameter, such as population mean is fixed but usually unknown. A statistic, such as changes with the random sample. The sampling distribution concerns that changing statistic.
Center and spread of a sample mean
For independent observations from a population with mean and finite standard deviation
The sample mean is unbiased because its sampling-distribution center equals the population mean. The standard deviation of the sampling distribution is often called the standard error when estimated from sample data.
For sampling without replacement, independence is only approximate. The usual 10% condition, keeps the finite-population dependence small enough for the familiar formula without a correction.
The Central Limit Theorem concerns shape
For sufficiently large independent samples from a population with finite variance, the sampling distribution of the mean is approximately normal, even if the population itself is not normal. Strong skewness and extreme outliers may require a larger sample than mild asymmetry.
If the population is normal, the sample mean is normal for any sample size. A common guideline such as n at least 30 is not a universal guarantee for every population. The theorem changes the distribution of sample means, not the original population's shape.
Example: calculate a probability for a mean
Suppose individual package weights have mean 500 grams and standard deviation 12 grams. For independent samples of 36 packages, the sample mean has mean 500 and standard deviation grams.
If a normal approximation is appropriate, the probability that the mean exceeds 504 grams is approximately because
Using 12 instead of 2 in that calculation would answer a question about one package, not a mean of 36 packages. The wording “sample mean” determines the relevant variability.
Sampling distributions of proportions
For a random sample with success probability , the sample proportion has mean and standard deviation under independence. A normal approximation is reasonable when expected successes np and expected failures are both sufficiently large, commonly at least 10.
If and the standard deviation is Repeated sample proportions tend to vary by a few percentage points around 0.4, even though the population proportion is fixed.
The large-count condition uses the population when studying a specified sampling distribution. In inference, the available value differs between a confidence interval and a null-hypothesis test, so the condition must match the procedure.
Example: understand the effect of sample size
To cut the standard deviation of a sample mean in half, increase by a factor of four. The square root in the denominator gives
Doubling the sample size reduces spread by a factor of not one half. Increasing improves precision but does not correct a biased sampling process. An unrepresentative sample can estimate the wrong target very precisely.
Two normal sampling distributions centered at 500: sample size 36 has standard deviation 2, while sample size 144 has standard deviation 1.
Larger samples make the sample mean less variable. Both sampling distributions remain centered at the same population mean.
Describe this figure
Two normal sampling distributions centered at 500: sample size 36 has standard deviation 2, while sample size 144 has standard deviation 1.
Curves: ; .
Try it yourself
A population has mean 80 and standard deviation 15. For independent samples of size 25, what are the mean and standard deviation of the sample mean? Is normality automatic?
Answer: The mean is 80 and the standard deviation is Normality is exact if the population is normal; otherwise, whether a normal approximation is reasonable depends on the population shape and sample size. The center and spread formulas alone do not establish normal shape.
What to remember
Identify whether the random quantity is an individual value, a mean, or a proportion. Sampling distributions describe repeated statistics, and larger samples reduce their spread through a square-root relationship.