Standard Error vs Standard Deviation

The One-Line Answer

"Standard deviation describes the spread of individual measurements. Standard error of the mean describes the uncertainty in the average. In physics lab reports, you almost always want the standard error."

If you take n repeated measurements of the same quantity, you get two useful statistics. The standard deviation σ tells you how much individual measurements vary from one another. The standard error of the mean σ/√n tells you how well you know the average - which is what you actually report as your measurement result.

These are related by a single formula:

Standard Error (SE) = σ / √n

where σ is the standard deviation and n is the number of measurements. As you take more measurements, SE decreases - your estimate of the mean improves. σ does not decrease - the spread of individual measurements stays the same regardless of how many you take.

Definitions and Formulas

For a set of n measurements x₁, x₂, ..., xₙ with mean x̄:

x̄ = (1/n) · Σxᵢ

The sample standard deviation is:

σ = √[ (1/(n−1)) · Σ(xᵢ − x̄)² ]

The standard error of the mean is:

SE = σ / √n

What Each Statistic Actually Means

Standard Deviation σ

σ characterises the distribution of individual measurements. It answers the question: if I take one more measurement, how far from the mean is it likely to be?

For a normal distribution:

• 68% of measurements fall within ±1σ of the mean
• 95% fall within ±2σ
• 99.7% fall within ±3σ

σ is a property of the measurement process itself. Taking more measurements does not change σ - it just gives you a better estimate of it.

Standard Error SE

SE characterises the uncertainty in the mean. It answers the question: how well do I know the average of my measurements?

SE = σ/√n decreases as n increases - more measurements give a better estimate of the true mean. This is why repeated measurements improve precision.

SE is the uncertainty you propagate through your formula when you have taken multiple readings of a quantity and are reporting their mean as your measurement.

Worked Example - Period of a Pendulum

A student times 10 complete oscillations of a pendulum and records the following periods (in seconds):

1.842, 1.856, 1.831, 1.848, 1.839, 1.853, 1.845, 1.837, 1.851, 1.844

Step 1 - Compute the mean:

x̄ = (1.842 + 1.856 + ... + 1.844) / 10
x̄ = 18.446 / 10
x̄ = 1.8446 s

Step 2 - Compute the deviations from the mean:

(1.842 − 1.8446) = −0.0026
(1.856 − 1.8446) = +0.0114
(1.831 − 1.8446) = −0.0136
(1.848 − 1.8446) = +0.0034
(1.839 − 1.8446) = −0.0056
(1.853 − 1.8446) = +0.0084
(1.845 − 1.8446) = +0.0004
(1.837 − 1.8446) = −0.0076
(1.851 − 1.8446) = +0.0064
(1.844 − 1.8446) = −0.0006

Step 3 - Compute the sample standard deviation:

Σ(xᵢ − x̄)² = 0.0026² + 0.0114² + ... + 0.0006²
              = 0.000007 + 0.000130 + 0.000185 + 0.000012
                + 0.000031 + 0.000071 + 0.0000002 + 0.000058
                + 0.000041 + 0.0000004
              = 0.000535

σ = √(0.000535 / 9) = √0.0000594 = 0.00771 s

Step 4 - Compute the standard error:

SE = σ / √n = 0.00771 / √10 = 0.00771 / 3.162 = 0.00244 s

Step 5 - Report the result:

T = 1.845 ± 0.002 s

The uncertainty reported is the standard error (SE = 0.00244 s, rounded to 0.002 s), not the standard deviation (0.00771 s). Reporting σ = 0.008 s instead would overstate the uncertainty in the mean by a factor of √10 ≈ 3.2.

When to Use Standard Deviation vs Standard Error

Use Standard Deviation (σ) when:

- Describing how much individual measurements vary
- Characterising the natural variability of a population
- Checking whether a single new measurement is consistent with a distribution
- Reporting the spread of a dataset in a histogram or distribution plot
- The quantity of interest is a single measurement, not a mean

Use Standard Error (SE) when:

- Reporting the uncertainty in a mean value
- Propagating uncertainty through a formula (use SE as the uncertainty of each averaged quantity)
- Comparing a measured mean with an accepted value
- Constructing confidence intervals for a mean
- The quantity of interest is the average of multiple measurements - which is almost always the case in physics labs

How the Number of Measurements Affects Each Statistic

Understanding how n affects σ and SE is essential for planning experiments effectively.

n (measurements)	σ	SE = σ/√n	Improvement vs n=1
1	0.10	0.100	-
4	0.10	0.050	2× better
9	0.10	0.033	3× better
16	0.10	0.025	4× better
25	0.10	0.020	5× better
100	0.10	0.010	10× better

Notice the diminishing returns. Going from 1 to 4 measurements halves the uncertainty. Going from 4 to 16 halves it again. Each halving requires 4× as many measurements. This is why there is a practical limit to how much improvement averaging can achieve - at some point systematic errors dominate and more measurements provide no further benefit.

Using Standard Error in Uncertainty Propagation

When you have taken n repeated measurements of a quantity and want to propagate its uncertainty through a formula, use SE - not σ - as the uncertainty value Δx in the propagation formula.

For example, if you measure velocity v ten times and compute v̄ = 3.75 m/s with σ = 0.08 m/s:

SE = 0.08 / √10 = 0.025 m/s

Use Δv = 0.025 m/s (SE) in the propagation formula, not Δv = 0.08 m/s (σ).

If you then compute kinetic energy KE = ½mv²:

ΔKE/KE = √[ (Δm/m)² + (2·Δv/v)² ]

Use Δv = 0.025 m/s here. Using σ = 0.08 m/s instead would overestimate the uncertainty in KE by a factor of √10.

Common Mistakes to Avoid

Mistake 1: Reporting σ instead of SE for a mean value

This is the most frequent error. If you measured something n times and are reporting the mean, your uncertainty is SE = σ/√n. Reporting σ overstates the uncertainty by a factor of √n - making your result look n times worse than it actually is.

Mistake 2: Using n instead of n−1 for standard deviation

The sample standard deviation formula uses n−1 in the denominator (Bessel's correction). Using n underestimates the true standard deviation, especially for small samples. Most spreadsheet functions (STDEV in Excel, std() in Python with ddof=1) use n−1 by default - but always check.

Mistake 3: Thinking more measurements always helps

More measurements reduce SE - but only if random error is the dominant source of uncertainty. If systematic errors are larger than your random uncertainty, more measurements will not improve your result. Identify and correct systematic errors before investing time in extra measurements.

Mistake 4: Computing SE without enough measurements

SE = σ/√n requires a reliable estimate of σ. With only 2 or 3 measurements, your estimate of σ itself has huge uncertainty. Aim for at least 5 measurements before computing SE - and treat SE from small samples with appropriate caution.

Mistake 5: Confusing SE with the uncertainty in σ

SE is the uncertainty in the mean x̄, not the uncertainty in the standard deviation σ. The uncertainty in σ itself (how well you know σ) is approximately σ/√(2(n−1)) - a separate quantity rarely needed in undergraduate labs.

Quick Reference

Statistic	Symbol	Formula	Meaning	Use when
Mean	x̄	Σxᵢ/n	Average value	Always - report this as your measurement
Sample SD	σ	√(Σ(xᵢ−x̄)²/(n−1))	Spread of individual measurements	Describing variability
Standard Error	SE	σ/√n	Uncertainty in the mean	Reporting uncertainty in a mean value

Frequently Asked Questions

Q1Should I report standard deviation or standard error in my physics lab report?

Standard error (SE = σ/√n) in almost all cases. You are reporting a mean value - SE is the uncertainty in that mean. Standard deviation describes the spread of individual measurements, which is a different (and less useful) quantity for most lab report purposes.

Q2What if I only took one measurement?

SE does not apply - you need multiple measurements to compute it. For a single measurement, your uncertainty comes from the instrument resolution (typically half the smallest scale division) or the manufacturer's stated accuracy. Take repeated measurements whenever possible.

Q3How many measurements do I need for SE to be meaningful?

At least 5, ideally 10 or more. With fewer than 5 measurements, your estimate of σ is unreliable, making SE unreliable too. The uncertainty in σ itself is approximately σ/√(2(n−1)), which is 50% of σ for n=2 but only 16% of σ for n=20.

Q4My standard deviation is larger than my standard error. Is that normal?

Yes, always. SE = σ/√n is always smaller than σ by a factor of √n. A mean value is always known more precisely than individual measurements. If SE > σ, you have made a calculation error.

Q5Can I use standard deviation as a conservative uncertainty estimate?

Yes, but you should state clearly that you are doing so and why. Using σ instead of SE is a deliberately conservative choice that overstates your uncertainty. It is acceptable in some contexts (e.g. when you are uncertain about systematic errors) but should not be the default.

Q6What is the difference between standard error and confidence interval?

A confidence interval is constructed from SE. The 95% confidence interval for a mean is approximately x̄ ± 2·SE (for large n). Reporting x̄ ± SE corresponds roughly to a 68% confidence interval. Some labs require 95% confidence intervals - check your lab guidelines.

Once you have computed your mean and standard error from repeated measurements, use the calculator to propagate SE through your formula - with full step-by-step working shown.

Open the Calculator →