Computing the Standard Error of a Sample Mean

From Sample Spread to Standard Error

In “The Form of a One-Sample t Interval,” you saw that the interval’s margin of error uses \(s/\sqrt{n}\). This tutorial focuses on calculating that quantity and understanding what it measures. The calculation is short, but it is important to keep the standard error separate from \(s\): they describe different sources of variability.

The sample standard deviation \(s\) describes the spread of individual observations in the sample around the sample mean \(\bar{x}\). The standard error describes the estimated spread of sample means from repeated samples of the same size. When the population standard deviation \(\sigma\) is unknown, \(s/\sqrt{n}\) estimates the standard deviation of the sampling distribution of \(\bar{x}\).

Definition: The estimated standard error of a sample mean is \(s/\sqrt{n}\), where \(s\) is the sample standard deviation and \(n\) is the sample size. It estimates how much sample means typically vary from sample to sample.

The denominator is the square root of the sample size, not the sample size itself. First find \(\sqrt{n}\), then divide \(s\) by that value. A larger sample generally gives a smaller standard error when the sample standard deviation is about the same. This does not mean the individual observations have become less variable; it means the mean is estimated with less sampling variability.

SE_{\bar{x}}=\frac{s}{\sqrt{n}}

Both \(s\) and \(SE_{\bar{x}}\) have the same units as the original measurements. The standard error is often smaller than \(s\), but it is not automatically smaller in every possible data set: it depends on both the observed \(s\) and \(n\). It is not the standard deviation of the individual observations, and it does not describe the range containing most individual values.

In “Standard Deviation of the Sample Mean,” the standard deviation of the sampling distribution was written as \(\sigma/\sqrt{n}\) when \(\sigma\) is known. In a one-sample t interval, \(\sigma\) is unknown, so we use \(s\) to estimate it. Therefore, \(s/\sqrt{n}\) is an estimated standard error, not the exact standard deviation of the sampling distribution.

Calculating and Checking the Standard Error

A reliable calculation has three parts: identify the sample standard deviation and sample size, divide \(s\) by \(\sqrt{n}\), and report the result with the original measurement units. Keep extra digits in intermediate work, then round the final value consistently with the context.

Identify \(s\) and \(n\).
Use the sample standard deviation and the number of observations. Do not substitute \(\bar{x}\) or \(n-1\) into the standard-error formula.

Find \(\sqrt{n}\).
Take the square root of the sample size before dividing.

Calculate \(s/\sqrt{n}\).
Show the substitution and give the result in the measurement units.

Say what the result describes.
It estimates the typical sample-to-sample variation in \(\bar{x}\), not the spread of individual observations.

The arithmetic can be checked by multiplying the calculated standard error by \(\sqrt{n}\): the result should be approximately \(s\), allowing for rounding. This check will not replace a careful interpretation, but it can help catch an incorrect square root or division.

The formula can be calculated whenever \(s\) and \(n\) are known. However, using it to support inference about a population mean also requires appropriate sampling and independence reasoning. As discussed in “The Form of a One-Sample t Interval,” check that data come from a random sample when generalizing to a population, and check independence. For sampling without replacement, use the 10% condition. A random assignment by itself supports conclusions about treatment effects; it does not by itself justify generalizing to a population. For a t interval, also consider whether the sample data are sufficiently well behaved for the procedure, especially when the sample is small.

Worked Examples

Worked Example: Standard Error of Adult Learners’ Study Time

A random sample of 64 adult learners reports weekly study time. The sample standard deviation is 12.0 hours. Calculate the estimated standard error of the sample mean study time.

State. The sample standard deviation, \(s=12.0\) hours, describes the spread of individual weekly study times in the sample. The sample size is \(n=64\). We want to estimate the sample-to-sample variability of \(\bar{x}\).

Plan. Use \(s/\sqrt{n}\). The data come from a random sample, which supports generalizing to the population represented by that sample. If the sample was taken without replacement, the 10% condition requires a population of at least 640 adult learners, since \(64\leq0.10(640)\). Independence is then reasonable under that condition. These checks matter for using the standard error in population inference; they do not change the arithmetic.

Do. Substitute the sample statistics:

SE_{\bar{x}}=\frac{s}{\sqrt{n}} =\frac{12.0}{\sqrt{64}} =\frac{12.0}{8} =1.5\text{ hours}.

Check the calculation by multiplying back: \(1.5(8)=12.0\) hours, which recovers \(s\). The estimated standard error is 1.5 hours.

Conclude. For repeated random samples of 64 learners, sample means of weekly study time would typically vary by about 1.5 hours from the population mean. The 12.0-hour sample standard deviation instead describes variation among individual learners in this sample.

Worked Example: A Small Sample of Reusable Bottles

A quality-control technician selects a random sample of 16 reusable bottles from a large production run. The standard deviation of the bottle capacities is \(s=6.0\) milliliters. Calculate the estimated standard error of the sample mean capacity.

Here, \(n=16\) and \(s=6.0\) milliliters. Apply the formula:

SE_{\bar{x}}=\frac{6.0}{\sqrt{16}} =\frac{6.0}{4} =1.5\text{ milliliters}.

As a check, \(1.5(4)=6.0\) milliliters. The estimated standard error is 1.5 milliliters. It is not the spread of capacities among individual bottles; that spread in the sample is represented by \(s=6.0\) milliliters.

The random selection supports using the sample to estimate the mean capacity for the production run. Because sampling is without replacement, the 10% condition is satisfied if the run contains at least 160 bottles. Independence is reasonable if that condition holds. Since this example has a small sample, a t interval would also require checking the sample data for strong skewness or outliers; that shape check is separate from calculating the standard error.

Worked Example: A Sample Size That Is Not a Perfect Square

A random sample of 50 homes is used to estimate mean daily electricity use. The sample standard deviation is \(s=14.0\) kilowatt-hours. Find the estimated standard error, rounded to four decimal places.

First calculate the square root and then divide:

SE_{\bar{x}}=\frac{14.0}{\sqrt{50}} =\frac{14.0}{7.0711} \approx1.9799\text{ kilowatt-hours}.

A check using the rounded result gives \(1.9799(7.0711)\approx14.000\) kilowatt-hours, which agrees with the sample standard deviation up to rounding. Thus the estimated standard error is approximately 1.9799 kilowatt-hours.

This is an estimate of the typical variability among sample means from repeated samples of 50 homes. It is not a claim that individual homes’ daily electricity use is usually within about 1.9799 kilowatt-hours of the sample mean. The individual-use spread is described by \(s=14.0\) kilowatt-hours.

For population inference, the random sample supports generalizing to the population of homes it represents. If homes were sampled without replacement, the 10% condition requires at least 500 homes in that population. Independence is reasonable if the condition is met. Before applying a t interval, also inspect the sample distribution for unusual skewness or outliers.

Worked Example: Comparing Two Sample Sizes

Suppose two random samples from a population have the same sample standard deviation, \(s=8.0\) minutes. One sample has \(n=25\) observations and the other has \(n=100\). Calculate and compare their estimated standard errors.

For the sample of 25 observations:

SE_{\bar{x}}=\frac{8.0}{\sqrt{25}} =\frac{8.0}{5} =1.6\text{ minutes}.

The check is \(1.6(5)=8.0\). For the sample of 100 observations:

SE_{\bar{x}}=\frac{8.0}{\sqrt{100}} =\frac{8.0}{10} =0.8\text{ minutes}.

The check is \(0.8(10)=8.0\). The larger sample has half the estimated standard error: \(0.8/1.6=0.5\). This follows because increasing the sample size from 25 to 100 multiplies \(n\) by 4, and \(\sqrt{n}\) by 2. With the same \(s\), doubling the square-root denominator halves the standard error. Neither calculation says the sample standard deviation has changed; both samples have \(s=8.0\) minutes.

Common Mistakes and AP Exam Tips

Reporting \(s\) as the standard error. If \(s=12\) and \(n=64\), the standard error is \(12/\sqrt{64}=1.5\), not 12. State which quantity describes individual observations and which estimates sample-mean variability.
Dividing by \(n\) instead of \(\sqrt{n}\). The formula is \(s/\sqrt{n}\). For example, when \(n=64\), divide by 8, not by 64.
Using the wrong statistic. The formula requires the sample standard deviation \(s\), not the sample mean \(\bar{x}\). In a one-sample t procedure, \(\sigma\) is unknown and \(s\) is used to estimate it.
Dropping the units. Standard error has the same units as the original measurements. A standard error of 0.8 minutes is not unit-free.
Describing the standard error as a range for individual observations. It estimates variability in sample means across repeated samples, not the spread of individual data values around \(\bar{x}\).
Claiming random assignment justifies generalizing to a population. Random sampling supports population generalization. Random assignment supports causal conclusions about treatment effects, but does not by itself make a sample representative of a population.
Rounding too early or omitting the calculation. Show \(s/\sqrt{n}\), substitute the values, and report a suitably rounded result. Keeping extra digits until the final step helps avoid small rounding inconsistencies.

A strong AP response identifies \(s\) and \(n\), shows the division by \(\sqrt{n}\), includes units, and explains what the result estimates. When the standard error will be used for inference about a population mean, also give the relevant sampling and independence reasoning rather than treating the calculation alone as justification for a procedure.

Key takeaway: For a sample mean, the estimated standard error is \(s/\sqrt{n}\). The sample standard deviation \(s\) describes the spread of individual observations; the standard error estimates the sample-to-sample spread of means. Both use the original measurement units.

Check Your Understanding

For each question, distinguish the spread of individual observations from the estimated variability of sample means.

A sample has \(n=36\) and \(s=9\) centimeters. Calculate the estimated standard error and show the square root used.
A researcher reports \(s=5.4\) kilograms and \(SE_{\bar{x}}=1.8\) kilograms. What is the sample size?
In a sample of 100 plants, the standard error of mean height is 0.7 centimeters. What does this quantity estimate, and what does it not describe?
Two samples have the same \(s\), but one has four times the sample size of the other. How do their estimated standard errors compare?
Why does random assignment alone not justify generalizing an estimated mean to a population?