How to Calculate Standard Error

You collect a sample, calculate its average, and get a result such as 52.4.

But that immediately creates another question: how precise is that average?

If you collected another random sample from the same population, you probably wouldn’t get exactly 52.4 again. You might get 51.9, 52.8, or 53.1. Different samples contain different observations, so their means vary.

Standard error measures that sampling variability.

For a sample mean, the standard error is usually estimated as:

SE_x̄ = s/(√n)

where:

• s = sample standard deviation
• n = sample size

For example, if a sample has a standard deviation of 12 and contains 36 observations:

SE_x̄ = 12/(√36) = 12/6 = 2

The standard error of the mean is 2.

The calculation is simple. Understanding what that 2 actually represents is the more important part.

What Standard Error Actually Measures

Imagine that instead of collecting one sample from a population, you could repeat the process thousands of times.

Each time, you:

1. randomly select a sample of the same size
2. calculate its mean
3. record that mean

The individual samples would contain different observations, so their means would not be identical.

You would end up with a distribution of sample means.

                 Population
                     │
        ┌────────────┼────────────┐
        ▼            ▼            ▼
     Sample 1     Sample 2     Sample 3
        │            │            │
        ▼            ▼            ▼
      Mean 1       Mean 2       Mean 3
        │            │            │
        └────────────┼────────────┘
                     ▼
             Distribution of
               Sample Means
                     │
                     ▼
              Standard Error

The standard deviation of that sampling distribution is the standard error of the mean.

This is why standard error is different from standard deviation.

Standard deviation describes how individual observations vary.

Standard error describes how an estimate, such as the sample mean, would vary across repeated samples.

Standard Error Formula

When the population standard deviation σ is known, the standard error of the sample mean is:

SE_x̄ = (σ)/(√n)

where:

• σ = population standard deviation
• n = sample size

In real-world analysis, however, the population standard deviation is often unknown.

Instead, you estimate it using the sample standard deviation s:

SE_x̄ = s/(√n)

where:

• s = sample standard deviation
• n = sample size

This second formula is the one you will commonly use when working with sample data.

The distinction between σ and s matters:

Population SD:  σ
Sample SD:      s

Known σ:        SE = σ / √n
Estimated σ:    SE ≈ s / √n

The sample standard deviation itself uses n-1 when calculated from raw sample data. You do not replace √n with √(n-1) in the standard-error formula.

The National Institute of Standards and Technology (NIST) similarly distinguishes between calculations using a known population standard deviation and calculations where the standard deviation must be estimated from sample measurements.

Source: NIST — Statistical Techniques

How to Calculate Standard Error

If you already know the sample standard deviation, calculating the standard error takes three steps.

Step 1: Find the Standard Deviation

Suppose:

Sample standard deviation (s) = 15
Sample size (n) = 100

Step 2: Take the Square Root of the Sample Size

√100 = 10

Step 3: Divide the Standard Deviation by √n

SE_x̄ = 15/10 = 1.5

So:

Standard error = 1.5

That tells us about the sampling variability of the sample mean, not the spread of the 100 individual observations.

Calculating Standard Error From Raw Data

Sometimes you are given the observations themselves rather than a standard deviation.

Consider this sample:

8, 10, 12, 14, 16

There are five observations:

n = 5

Step 1: Calculate the Sample Mean

x̄ = (8+10+12+14+16)/5

x̄ = 12

Step 2: Find Each Deviation From the Mean

Value    Deviation    Squared deviation
  8         -4               16
 10         -2                4
 12          0                0
 14          2                4
 16          4               16

The squared deviations add to:

16+4+0+4+16=40

Step 3: Calculate the Sample Standard Deviation

For a sample, the variance is:

s² = (Σ(x_i-x̄)²)/(n-1)

So:

s² = 40/4=10

and:

s=√10≈3.162

Step 4: Calculate the Standard Error

Now divide the sample standard deviation by the square root of the sample size:

SE_x̄ = 3.162/(√5)

SE_x̄ ≈ 3.162/2.236

SE_x̄ ≈ 1.414

The estimated standard error of the sample mean is therefore approximately:

1.41

Notice where n-1 appeared: it was used to calculate the sample variance.

Once we had s, the standard error was still:

s/(√n)

not:

s/(√(n-1))

Why Do We Divide by the Square Root of n?

The √n in the formula explains one of the most important properties of standard error: larger samples generally produce more precise estimates of the mean.

Suppose the standard deviation stays at 20.

With 25 observations:

SE = 20/(√25) = 4

With 100 observations:

SE = 20/(√100) = 2

With 400 observations:

SE = 20/(√400) = 1

So:

Sample size     Standard error
     25               4
    100               2
    400               1

Increasing the sample from 25 to 100 makes it four times as large, but only halves the standard error.

That’s because standard error decreases according to the square root of sample size, not sample size itself.

To cut the standard error in half while variability stays constant, you need approximately four times as many independent observations.

Standard Error vs Standard Deviation

Standard error and standard deviation are closely related, but they answer different questions.

	Standard Deviation	Standard Error of the Mean
Describes	Variation among observations	Sampling variability of the mean
Typical formula	s	s/√n
Changes with sample size	Not mechanically through a √n divisor	Decreases as n increases, all else equal
Main question	How spread out are the observations?	How precisely is the mean estimated?

Suppose 100 measurements have:

Mean = 50
SD   = 10

The standard deviation of 10 describes the variability among those measurements.

The standard error is:

SE = 10/(√100) = 1

That value of 1 describes the estimated sampling variability of the mean.

They describe different forms of uncertainty, so substituting one for the other changes the interpretation.

Standard Error Is Not Only About Means

The phrase “standard error” is sometimes used as though it means only the standard error of the mean.

More precisely, a standard error describes the sampling variability of an estimated statistic.

Statistics that can have standard errors include:

• sample means
• proportions
• regression coefficients
• differences between means
• differences between proportions
• other parameter estimates

Each estimator can have its own standard-error formula.

The familiar:

s/(√n)

is specifically the estimated standard error of the sample mean, often abbreviated SEM.

When someone asks how to calculate standard error without specifying an estimator, SEM is often what they mean.

Standard Error of a Proportion

For a sample proportion p̂, the estimated standard error is:

SE_p̂ = √((p̂(1-p̂))/n)

where:

• p̂ = sample proportion
• n = sample size

Suppose 60% of 400 surveyed customers choose a particular response.

Then:

p̂=0.60

and:

n=400

So:

SE_p̂= √((0.60(1-0.60))/400)

= √(0.24/400)

= √0.0006

≈0.0245

The estimated standard error of the proportion is approximately:

0.0245, or 2.45 percentage points.

This is another example of the same underlying idea: the standard error measures how much an estimate would tend to vary across repeated samples.

Standard Error and Confidence Intervals

Standard error is also an important part of many confidence intervals.

The general structure is:

Estimate ± (critical value × standard error)

The quantity:

critical value × standard error

is the margin of error.

For a population mean, the appropriate critical value depends on the situation.

When Population Standard Deviation Is Known

If σ is known and the conditions for a normal-based interval are satisfied, a confidence interval can use a z critical value:

x̄ ± z*(σ)/(√n)

For a 95% confidence level:

z* ≈ 1.96

When Population Standard Deviation Is Estimated

More commonly, σ is unknown and is estimated using the sample standard deviation s.

For the usual one-sample t interval:

x̄ ± t*s/(√n)

The t critical value depends on the confidence level and degrees of freedom.

When s is calculated from the same sample:

df=n-1

Penn State’s statistics materials give the same structure for a confidence interval for a population mean:

x̄± t*s/(√n)

Source: Penn State STAT 200 — Confidence Intervals

This is why 1.96 × SE is not a universal formula for every 95% confidence interval.

The correct multiplier depends on the statistical procedure and its assumptions.

Example: Standard Error and a 95% Confidence Interval

Suppose a sample contains:

n = 25
Mean = 80
Sample SD = 10

First calculate the standard error:

SE = 10/(√25)=2

Because the population standard deviation is being estimated from the sample, a one-sample t interval uses:

df=25-1=24

For a two-sided 95% confidence interval with 24 degrees of freedom, the critical value is approximately:

t*=2.064

The margin of error is therefore:

2.064×2=4.128

So the interval is:

80±4.128

or approximately:

75.87 to 84.13

The standard error is 2.

The margin of error is approximately 4.13.

The confidence interval is approximately 75.87 to 84.13.

These quantities are related, but they are not interchangeable.

Does Standard Error Require Normally Distributed Data?

Not in the simple sense that “standard error requires normal data.”

The formula:

SE_x̄=(σ)/(√n)

describes the standard deviation of the sample mean under the usual independent-sampling setup when observations have finite variance.

Normality becomes especially relevant when deciding which inferential procedure or distributional approximation to use.

If the underlying population is normal, the sampling distribution of the mean is normal.

For non-normal populations, the central limit theorem tells us that under suitable conditions the sampling distribution of the mean becomes approximately normal as sample size increases.

That does not mean there is one universal sample-size cutoff that makes every dataset safe to treat as normal. Strong skew, heavy tails, outliers, dependence between observations, and the sampling design can all matter.

So rather than asking only:

“Is my raw data normal?”

also ask:

• Were observations sampled appropriately?
• Are observations independent, or has dependence been modeled?
• Are there severe outliers or heavy tails?
• Is the sample large enough for the approximation being used?
• Is the standard-error formula appropriate for the sampling or study design?

The assumptions belong to the statistical method, not to the idea of standard error alone.

A Small Standard Error Does Not Mean the Estimate Is Correct

A small standard error means an estimator has low estimated sampling variability under the model and sampling procedure being used.

It does not automatically mean the estimate is unbiased or that the study is well designed.

Imagine surveying 100,000 people about an election but collecting responses only from one highly unrepresentative group.

The enormous sample could produce a very small conventional standard error.

The estimate could still be systematically wrong because the sampling process is biased.

Standard error captures sampling variability. It does not automatically account for problems such as:

• selection bias
• nonresponse bias
• systematic measurement error
• confounding
• incorrect model assumptions
• dependence ignored by the calculation

A precise estimate can still be a biased estimate.

How Sample Size Affects Standard Error

For the sample mean:

SE=(σ)/(√n)

If variability stays constant, increasing n reduces the standard error.

But the improvement has diminishing returns.

If you double the sample size:

1/(√2)≈0.707

So the new standard error is about 70.7% of the original, a reduction of roughly 29.3%.

If you quadruple the sample size:

1/(√4) = 1/2

the standard error is cut in half.

This relationship is important when planning studies because increasingly large improvements in precision require increasingly large samples.

Common Standard Error Mistakes

Using √(n−1) in the SEM Formula

For the usual estimated standard error of a sample mean:

SE_x̄=s/(√n)

The n-1 correction is already involved when s is calculated as the sample standard deviation.

Confusing SD and SE

Standard deviation describes the spread of observations.

Standard error describes the sampling variability of an estimator.

They answer different questions.

Treating 1.96 as the Universal 95% Multiplier

A z critical value of approximately 1.96 is appropriate in certain normal-based situations.

When population standard deviation is estimated for a one-sample mean, a t critical value is commonly used instead.

Interpreting SE as ± Uncertainty Around One Observation

If:

Mean = 50
SE = 2

that does not mean individual observations are generally between 48 and 52.

SE concerns the estimated mean, not the spread of individual values.

Assuming a Small SE Eliminates Bias

Increasing sample size can reduce sampling variability.

It does not automatically correct a biased sampling method or systematic measurement error.

Ignoring the Sampling Design

The simple s/√n calculation assumes a straightforward independent-sampling setting.

Clustered observations, repeated measurements, complex survey designs, time-series dependence, and other structures may require different standard-error calculations.

Should You Report Standard Error or Standard Deviation?

That depends on what you are trying to communicate.

Use standard deviation when you want to describe variability among the observations.

Use standard error when you want to describe the estimated sampling precision of a statistic.

A confidence interval is often more directly interpretable when communicating the plausible range of a population parameter.

For example:

Mean = 50
SD = 8

describes the observations differently from:

Mean = 50
SE = 2

and differently again from:

95% CI = 45.9 to 54.1

The correct choice depends on whether the purpose is descriptive statistics, statistical inference, or communication of uncertainty.

Frequently Asked Questions

What is the formula for standard error?

For the estimated standard error of a sample mean:

SE_x̄=s/(√n)

where s is the sample standard deviation and n is the sample size.

If the population standard deviation σ is actually known:

SE_x̄=(σ)/(√n)

What does standard error tell you?

Standard error describes the sampling variability of an estimate.

For a sample mean, it tells you how much sample means would tend to vary if the sampling process were repeated many times under the same conditions.

Is standard error the same as standard deviation?

No.

Standard deviation describes variability among observations. Standard error describes sampling variability of an estimate.

Why does standard error decrease as sample size increases?

For a sample mean, standard error is proportional to:

1/(√n)

As the sample size grows, the denominator increases and the standard error decreases, assuming variability remains comparable.

Is standard error the same as margin of error?

No.

Standard error measures sampling variability.

Margin of error is typically:

critical value×standard error

So the standard error is one component used to calculate a margin of error.

Is standard error the same as a confidence interval?

No.

A standard error is a measure of sampling variability.

A confidence interval combines an estimate, a standard error, and an appropriate critical value:

estimate±(critical value× SE)

When do I use s instead of σ?

Use σ when the population standard deviation is genuinely known.

When it is unknown and you have sample data, you commonly estimate it using the sample standard deviation s.

Why isn’t the SEM s/√(n−1)?

Because the n-1 correction belongs to the calculation of the sample variance:

s²=(Σ(x_i-x̄)²)/(n-1)

Once s has been calculated, the estimated standard error of the mean is:

s/(√n)

Can the standard error of the mean be larger than the sample standard deviation?

For the ordinary formula:

SE_x̄=s/(√n)

and n>1, the SEM is smaller than s because √n>1.

This statement applies specifically to the standard error of the mean. Other statistics have different standard-error formulas and should not be compared with the raw-data standard deviation in the same way.

How can I reduce standard error?

For a sample mean, increasing the number of independent observations reduces the standard error when other conditions remain comparable.

Improving measurement quality and choosing an efficient study design may also improve precision, depending on the source of variability.

However, reducing standard error does not automatically eliminate bias or other study-design problems.

Quick Reference

For a sample mean when the population SD is known:

SE_x̄=(σ)/(√n)

When population SD is unknown and estimated from the sample:

SE_x̄=s/(√n)

For a sample proportion:

SE_p̂=√((p̂(1-p̂))/n)

For a confidence interval:

estimate±(critical value× SE)

Remember the distinction:

Standard deviation
        │
        ▼
Spread of observations

Standard error
        │
        ▼
Sampling variability of an estimate

Confidence interval
        │
        ▼
Estimate ± critical value × SE

Related Calculator

For checking SEM from a standard deviation and sample size, use the Standard Error Calculator.

Final Thoughts

The formula for the standard error of a sample mean is easy to memorize:

SE_x̄=s/(√n)

But the formula makes more sense once you understand what is varying.

A random sample gives you one sample mean. A different random sample would usually give you a slightly different mean. If you repeatedly sampled from the same population, those sample means would form their own distribution.

Standard error describes the spread of that sampling distribution.

That explains why standard error is not the same as standard deviation, why larger samples generally produce smaller standard errors, and why standard error appears throughout confidence intervals and hypothesis tests.

The central distinction is simple:

Standard deviation describes variation in the data. Standard error describes sampling variation in an estimate.