Statistics › 🎯 Sampling & the CLT

Standard Error: How Precise Is Your Estimate?

Why a sample mean has its own spread — and how that spread shrinks as you collect more data.

Intro StatisticsAP Statistics levelCollege Stats 101
💡
The big idea: Every time you compute a statistic like a sample mean, you get a slightly different answer depending on which sample you happened to draw. The standard error measures how much that answer bounces around from sample to sample. For the mean it equals the population's standard deviation divided by the square root of the sample size, so bigger samples give steadier estimates — but only with diminishing returns.
🎯 By the end, you'll be able to
  • Distinguish the standard deviation (spread of the data) from the standard error (spread of an estimate)
  • State that the standard error of the sample mean is sigma divided by the square root of n
  • Explain why the standard error shrinks as the sample size n grows
  • Apply the 1/sqrt(n) rule, including the fact that halving the standard error takes four times the data
📎 You should already know
  • Mean and standard deviation
  • The sampling distribution of the mean

Your estimate has a spread of its own

Suppose the average income in a city is some fixed but unknown number. You draw a random sample of 50 people and compute their average income. Your neighbor draws a different random sample of 50 people and computes their average. You will not get exactly the same answer — and neither of you will land exactly on the true city-wide average.

That wobble is unavoidable. Each sample is a slightly different snapshot, so each sample mean is a slightly different estimate. The natural question is: how big is that wobble? If two honest samples can disagree by a mile, your single estimate is shaky. If they barely differ, your estimate is trustworthy. The standard error is the number that answers this question.

🔑 What standard error means
The standard error of a statistic is the standard deviation of that statistic across all the samples you could have drawn. It is not a property of the raw data — it is a property of your estimate. A small standard error means that different random samples would give you nearly the same answer, so your one sample is a precise estimate.

SD and SE are two different rulers

These two look almost identical on the page, so it is worth pinning down the difference:

  • The standard deviation \( \sigma \) measures how spread out the individual raw values are. Incomes range widely, so \( \sigma \) is large. This does not shrink when you collect more data — the population is just as spread out as it always was.
  • The standard error measures how spread out the sample means are. This does shrink as your samples get bigger, because averaging many values smooths out the extremes.

Same units, completely different jobs. One describes the data; the other describes the quality of your estimate.

The formula for the mean

For the special case of a sample mean, the standard error has a clean formula. It ties the spread of your estimate directly to the spread of the raw data and the size of your sample:

\[ \text{SE} = \frac{\sigma}{\sqrt{n}} \]
Standard error of the sample mean: the population standard deviation sigma divided by the square root of the sample size n.

Watch the standard error shrink

The simulation below draws sample after sample from a population and piles up their averages. The width of that bottom pile is the standard error — you can read it off directly.

Focus on the sample size n slider. Set n small and the sample means scatter widely: your estimate is imprecise. Raise n and the pile tightens around the true center. The readout Spread of means (SE) is the standard error measured from your draws, and the CLT prediction \( \sigma/\sqrt{n} \) is the theoretical value — watch them track each other as you change n.

🎮 Interactive: standard error as sample size grows LIVE
The bottom histogram collects sample means; its width is the standard error. Slide n upward and the spread of means shrinks toward zero, matching the sigma/sqrt(n) prediction shown in the readouts.
✨ Diminishing returns are built in
Because the formula divides by \( \sqrt{n} \) and not by \( n \), doubling your sample size does not halve the standard error. To cut the spread of your estimate in half, you need four times as much data. To make it a third as wide, you need nine times the data. This is why very precise estimates get expensive fast — each extra digit of precision costs disproportionately more sampling.
\[ \text{SE}_{\text{new}} = \frac{\sigma}{\sqrt{4n}} = \frac{1}{2}\cdot\frac{\sigma}{\sqrt{n}} = \tfrac{1}{2}\,\text{SE}_{\text{old}} \]
Quadrupling the sample size (n to 4n) pulls the square root of 4 = 2 out of the denominator, halving the standard error.
📝 Worked example: A population of adult heights has standard deviation sigma = 9 cm. You plan to estimate the average height from a random sample. (a) What is the standard error of the mean for a sample of n = 9? (b) For n = 81? (c) You want the standard error to be 1 cm — what sample size do you need?
  1. (a) SE = sigma / sqrt(n) = 9 / sqrt(9) = 9 / 3 = 3 cm.
  2. (b) SE = 9 / sqrt(81) = 9 / 9 = 1 cm. Note that n jumped by a factor of 9 (from 9 to 81) but the standard error only dropped by a factor of 3 (from 3 to 1) — that is the square root at work.
  3. (c) Set 9 / sqrt(n) = 1, so sqrt(n) = 9, giving n = 81. This matches part (b).
✓ (a) 3 cm, (b) 1 cm, (c) n = 81. Notice that individual heights still vary with sigma = 9 cm no matter how large the sample gets — only the standard error of the estimate shrinks.

Check your understanding

1. Which quantity shrinks as you collect a larger sample?
Sigma describes the raw data and does not change with sample size. The standard error, sigma/sqrt(n), gets smaller as n grows because its denominator grows.
2. A population has standard deviation sigma = 30. For a sample of size n = 9, the standard error of the mean is…
SE = sigma/sqrt(n) = 30/sqrt(9) = 30/3 = 10.
3. You currently sample n = 50 people. To HALVE the standard error of your estimated mean, you should sample about…
The standard error depends on sqrt(n), so halving it requires four times the data: 50 x 4 = 200.
4. What is the key difference between standard deviation and standard error?
The standard deviation describes how spread out the individual data values are; the standard error describes how much a statistic (like the sample mean) varies from sample to sample.
✅ Key takeaways
  • Any statistic computed from a sample, like the sample mean, would come out slightly differently for a different random sample.
  • The standard error measures that sample-to-sample wobble — it is the standard deviation of the estimate, not of the raw data.
  • For the sample mean, SE = sigma/sqrt(n): the population standard deviation divided by the square root of the sample size.
  • Standard deviation (sigma) describes the spread of the data and does not change with n; standard error shrinks as n grows.
  • Because of the square root, halving the standard error takes four times the data — precision has diminishing returns.