standarddeviationcalculator.net

Updated

Sampling and estimation calculators

Tools for working from a sample to the population it came from: how many people to survey, how precise an estimate is, and how a sample mean or proportion behaves from one sample to the next. The last group covers process control limits and data-quality checks.

Which calculator do I need?

You have or wantUse
How many responses you need for a target margin of errorSample size calculator
The ± figure for a poll result or a sample mean you already haveMargin of error calculator
A range that likely contains the population meanConfidence interval calculator
A range for the population standard deviation or varianceConfidence interval for SD
The standard error of the mean from raw data, or the SD behind a published SEStandard error calculator
The chance that a sample mean lands in a range, given μ, σ and nCentral limit theorem calculator
A binomial probability approximated with the normal curveNormal approximation calculator
Control chart limits for a production processUpper control limit calculator

Plan a sample

Decide the sample size, or find the precision a given sample size buys you.

Estimate a population value

Point estimates, standard errors and confidence intervals for a mean, a proportion or a standard deviation.

Sampling distributions

How x̄ and p̂ vary from sample to sample, and when the normal curve can stand in for the binomial.

Process control, forecasting and data checks

Keep a process within limits, project a growth rate forward, and test whether reported figures look natural.

Standard deviation, standard error and margin of error

These three are easy to confuse because each is built from the one before. The standard deviation describes how spread out individual values are. The standard error describes how much the sample mean would vary if you repeated the study: SE = s / √n, so it shrinks as the sample grows while the SD does not. The margin of error is the standard error multiplied by a critical value (1.96 for 95% with a large sample), and a confidence interval is the estimate plus or minus that margin.

A worked comparison

A poll finds 50% support. With 1,000 respondents the 95% margin of error is 1.96 × √(0.5 × 0.5 / 1000) = 0.031, or ±3.1 percentage points. With 400 respondents it is 1.96 × √(0.25 / 400) = ±4.9 points. Going the other way, the sample size calculator shows that ±3 points needs 1.96² × 0.25 / 0.03² = 1,067.1, rounded up to 1,068 people. Precision improves with the square root of n: halving the margin of error takes four times the sample.

Where the normal curve comes from

The central limit theorem says the mean of a large enough sample is close to normally distributed, whatever the shape of the population, with standard deviation σ / √n. That is why confidence intervals and z-tests work on skewed data. The same idea lets you treat a binomial count as normal once np and n(1 − p) are both at least 10; the normal approximation and continuity correction calculators show how close the approximation gets to the exact answer.

Common mix-ups

  • Reporting SE as if it were SD. Error bars labelled ± SE look much tighter than ± SD. To recover the SD, multiply by √n.
  • Using the 1.96 critical value for small samples. For a mean with n below about 30, the t critical value is larger; the confidence interval calculator uses t for this reason.
  • Forgetting the finite population correction. When the sample is more than about 5% of the population, the correction noticeably shrinks the standard error and the required sample size.

Guides to read alongside

Common questions

Does a larger population need a larger sample?

Barely, once the population is large. A survey of 1,068 people gives about ±3 points at 95% confidence whether the population is 100,000 or 100 million. Population size only matters when the sample is a sizeable share of it, which the finite population correction handles.

What does a 95% confidence interval mean?

If you repeated the sampling many times and built an interval each time, about 95% of those intervals would contain the true population value. It does not mean there is a 95% probability that this particular interval contains it; the interval either does or does not.

Why does the sample proportion need np ≥ 10?

The normal approximation to the sampling distribution of p̂ breaks down when successes or failures are rare, because the true distribution is then lopsided. Requiring at least 10 expected successes and 10 expected failures keeps it close enough to symmetric for normal-based intervals and tests.