Interpreting Standard Error for Means

What Does the Standard Error Describe?

A sample mean is not fixed in advance: if a different random sample were drawn from the same population, its mean would usually be different. The standard error describes the spread of those sample means over repeated samples of the same size, taken under the same sampling conditions. It is a measure of how much the sample mean varies from sample to sample.

This is different from the population standard deviation. The population standard deviation, \(\sigma\), describes the spread of individual observations in the population. The standard error of the sample mean describes the spread of averages computed from samples. Both are measurements of variability, but they describe different quantities.

The formula \(\sigma/\sqrt{n}\) assumes that the observations are independent and come from the same population with finite standard deviation \(\sigma\). In this setting, \(n\) is the number of observations in each sample. The formula gives the standard deviation of the sample mean's sampling distribution; it does not require that the population itself be normally distributed.

The standard error has the same units as the variable being measured. If the observations are measured in millimeters, then \(\sigma\) is in millimeters and \(\sigma/\sqrt{n}\) is also in millimeters. Although the variance of the sample mean is measured in squared units, taking its square root returns to the original measurement units:

This is a units check as well as a useful interpretation: the standard error is a standard deviation, not a variance. Its role is to describe the typical scale of variation among repeated sample means. It does not say that every sample mean will differ from the population mean by exactly one standard error.

Calculating and Interpreting the Standard Error

When the population standard deviation is known, use \(\sigma/\sqrt{n}\). To interpret the result, name the statistic whose spread it describes, identify the repeated-sampling setting, and include the measurement units. A standard error of 2 minutes, for example, describes the spread of sample means in minutes, not the spread of individual times.

A standard error can be interpreted even when the shape of the sampling distribution has not been established. It gives the standard deviation of that distribution, but by itself it does not imply that most sample means fall within any particular number of standard errors of the population mean. Such a probability statement would require additional information about the distribution's shape.

When the Population Standard Deviation Is Unknown

In many applications, \(\sigma\) is not known. A sample's standard deviation \(s\) can then be used to estimate the standard error. The resulting quantity, \(s/\sqrt{n}\), is an estimated standard error, rather than the exact standard error calculated from the population parameter. The sample standard deviation describes variation among the individual observations in that sample; dividing it by \(\sqrt{n}\) estimates the spread of sample means.

This calculation uses the same measurement units for \(s\) and for the estimated standard error. Its interpretation should make clear that the population standard deviation was not known and that the reported standard error is an estimate.

Why the Distinction Matters

Standard error is useful because many statistical questions concern a sample statistic, such as \(\bar{x}\), rather than an individual observation. Describing the variability of that statistic requires a measure of the statistic's sampling distribution. The population standard deviation alone does not express that spread: it describes individuals. The standard error puts the sample mean's variability on a scale appropriate to the statistic.

The units can help diagnose a mistake. If a study measures cholesterol concentration in milligrams per deciliter, then \(\sigma\), \(s\), and the standard error of the mean are all expressed in milligrams per deciliter. A variance would instead be expressed in squared units. Also check what quantity a reported number describes: a value can be a standard deviation of individual measurements or a standard error of a statistic, but those labels are not interchangeable.

Common Mistakes

• Confusing standard error with population standard deviation. The mistaken reasoning is that both are measures of spread, so they describe the same spread. The correction is to identify what varies: \(\sigma\) describes individual observations, while the standard error describes sample means.
• Dividing by \(n\) instead of \(\sqrt{n}\). The standard error formula uses the square root of the sample size: \(SE(\bar{x})=\sigma/\sqrt{n}\), not \(\sigma/n\).
• Calling \(s/\sqrt{n}\) the exact standard error when \(\sigma\) is unknown. It is an estimate based on the sample standard deviation. State that it is estimated.
• Assuming a standard error guarantees a particular percentage of sample means nearby. A standard deviation alone does not establish such a percentage. A statement about probabilities also needs information about the sampling distribution's shape.