What Is a Confidence Interval?
A confidence interval is a range of values, built around a sample statistic, that likely contains the true population value. To calculate one, take your sample statistic (a mean or a proportion), then add and subtract a margin of error: critical value × standard error. For a 95% confidence interval on a mean, the formula is x̄ ± t × (s ÷ √n), where x̄ is the sample mean, s is the sample standard deviation, and n is the sample size. Plug your numbers into the confidence interval calculator to get the full interval, margin of error, and critical value in seconds.
Confidence intervals show up everywhere sample data is used to estimate something about a larger group: political polling (“52% ± 3 points”), A/B testing, clinical trials, quality control, and classroom statistics homework.
The Confidence Interval Formula, Step by Step
There are two common versions, depending on whether you’re estimating a mean (an average) or a proportion (a percentage).
For a Sample Mean
- Find your sample mean (x̄) and sample standard deviation (s). If you’re starting from a raw list of numbers, the standard deviation calculator computes both instantly.
- Find the standard error: SE = s ÷ √n, where n is your sample size.
- Look up the critical value (t) for your confidence level and degrees of freedom (n − 1). This calculator does that lookup for you.
- Calculate the margin of error: margin = t × SE.
- Build the interval: x̄ − margin to x̄ + margin.
For a Sample Proportion
- Find your sample proportion (p̂) — the percentage of your sample with the trait you’re measuring (e.g. 62% said “yes”).
- Find the standard error: SE = √(p̂(1 − p̂) ÷ n).
- Look up the critical value (z) for your confidence level — 1.960 for 95%, 2.576 for 99%, etc.
- Calculate the margin of error: margin = z × SE.
- Build the interval: p̂ − margin to p̂ + margin.
Worked Example: Mean
A researcher measures the resting heart rate of 25 adults. The sample mean is 72 bpm, with a sample standard deviation of 8 bpm. What’s the 95% confidence interval?
- Standard error: SE = 8 ÷ √25 = 1.6
- Degrees of freedom: 25 − 1 = 24, so the t critical value for 95% confidence is 2.064
- Margin of error: 2.064 × 1.6 = 3.30
- 95% CI: 72 − 3.30 to 72 + 3.30 = 68.70 to 75.30 bpm
You can be 95% confident the true average resting heart rate of the population this sample was drawn from falls between 68.70 and 75.30 bpm.
Worked Example: Proportion
A survey of 400 shoppers finds that 62% prefer online returns over in-store returns. What’s the 90% confidence interval?
- Standard error: SE = √(0.62 × 0.38 ÷ 400) = 0.0243
- Critical value for 90% confidence: z = 1.645
- Margin of error: 1.645 × 0.0243 = 0.04 (4 percentage points)
- 90% CI: 62% − 4 to 62% + 4 = 58% to 66%
Comparing Confidence Levels
The confidence level you choose directly controls how wide the interval is — higher confidence means a wider, less precise range.
| Confidence Level | Critical Value (z, large n) | Relative Width vs. 95% | Typical Use |
|---|---|---|---|
| 80% | 1.282 | Narrower | Quick exploratory estimates |
| 90% | 1.645 | ~16% narrower | Business/market research |
| 95% | 1.960 | Baseline | Most scientific research, polling |
| 98% | 2.326 | ~19% wider | Higher-stakes decisions |
| 99% | 2.576 | ~31% wider | Medical/clinical, quality control |
Notice that going from 95% to 99% confidence doesn’t just add a little precision — it makes the interval about 31% wider for the same data. There’s no free lunch: more confidence requires either a wider interval or a bigger sample.
Common misconception: A “95% confidence interval” does NOT mean there’s a 95% chance the true value falls inside this specific interval. It means that if you repeated the sampling process many times and built an interval each time, about 95% of those intervals would contain the true value. The interval you calculated either does or doesn’t contain it — you just don’t know which.
Z-Score vs. T-Score: Which Should You Use?
This trips up a lot of students and analysts:
- Use the t-distribution when you only know your sample standard deviation (the normal case for a mean) — it’s wider than the z-distribution to account for the extra uncertainty of estimating spread from a limited sample. The smaller your sample, the wider the t-distribution relative to z.
- Use the z-distribution when you know the true population standard deviation (rare in practice), or when calculating a confidence interval for a proportion, since the standard error formula there doesn’t rely on an estimated standard deviation the same way.
- As sample size grows (roughly n ≥ 120), the t-distribution converges to the z-distribution, so the difference becomes negligible.
If you’re not sure which value your data implies, the confidence interval calculator handles this automatically based on which mode you select.
How to Get a Narrower Confidence Interval
A narrower interval means a more precise estimate. Three levers control the width:
- Increase your sample size. The standard error shrinks with the square root of n, so quadrupling your sample size only halves your margin of error — but it’s the most reliable way to tighten an interval.
- Reduce variability in your data. More consistent measurement methods lower your standard deviation, which directly narrows the interval.
- Accept a lower confidence level. Dropping from 99% to 90% confidence narrows the interval, but you’re trading certainty for precision — decide which matters more for your use case.
Confidence Intervals in Excel or Google Sheets
Both tools have built-in functions:
- For a mean:
=CONFIDENCE.T(alpha, standard_dev, size)returns the margin of error using the t-distribution (Excel). In Sheets,=CONFIDENCE.T()works the same way. - Manual approach: Compute the standard error with
=STDEV(range)/SQRT(COUNT(range)), then multiply by the critical t-value from a table or=T.INV.2T(1-confidence, degrees_of_freedom). - For a proportion: There’s no dedicated built-in function — compute SE manually with
=SQRT(p*(1-p)/n)and multiply by the z critical value (=NORM.S.INV(1-(1-confidence)/2)).
Frequently Asked Questions
What sample size do I need for a reliable confidence interval? For a mean, there’s no strict minimum, but the t-distribution correction matters most when n is small (under 30). For a proportion, a common rule of thumb is that both n × p̂ and n × (1 − p̂) should be at least 5 (some sources say 10) for the normal approximation to hold up well.
What’s the difference between confidence interval and margin of error? The margin of error is just the ± amount (critical value × standard error). The confidence interval is the full range you get after adding and subtracting that margin from your sample statistic.
Can a confidence interval be one-sided? Yes — everything above assumes a standard two-sided (two-tailed) interval, which is by far the most common. One-sided intervals use a different critical value and are used when you only care about a lower or upper bound (e.g. “at least X%”).
Does a wider confidence interval mean my study is bad? Not necessarily. A wide interval usually just reflects a small sample size or highly variable data — it’s honest about the uncertainty rather than a sign of a flawed study. A narrow interval from a biased sample can actually be more misleading.
The Bottom Line
A confidence interval answers a precise question: given my sample, what range likely contains the true population value, at a stated confidence level? For a mean, that’s x̄ ± t × (s ÷ √n); for a proportion, it’s p̂ ± z × √(p̂(1 − p̂) ÷ n). Both formulas take seconds by hand once you have your sample statistics — or skip the lookup tables entirely with the confidence interval calculator, which shows the interval, margin of error, and critical value together. If you’re starting from raw data, run it through the standard deviation calculator first, and pair your result with the z-score calculator if you also need to convert individual values to percentiles.