A confidence interval (CI) is a range of values that likely contains the true population parameter. When you survey or measure a sample, your sample mean is your best single estimate — but a single number can't capture the inherent uncertainty in any sample. A confidence interval adds that uncertainty explicitly, giving you an upper and lower bound around your estimate.
If you measure the average running pace of 50 marathon runners and get a sample mean of 5:30 min/km with a 95% confidence interval of [5:15, 5:45], it means: if you repeated this study many times with different random samples, 95% of those intervals would contain the true population mean.
Key components:
Wider intervals indicate more uncertainty; narrower intervals indicate more precision. The width depends on your sample size, standard deviation, and chosen confidence level. Increasing your sample size is the most effective way to narrow a confidence interval.
For a population mean using a sample (the most common case), the confidence interval is:
CI = x̄ ± Z × (s / √n)
Where:
When to use t instead of Z: When your sample size is small (n < 30) and/or the population standard deviation is unknown (which is almost always), you should technically use the t-distribution instead of Z. The t-distribution has heavier tails, producing wider (more conservative) intervals for small samples. For n ≥ 30, the difference is negligible. Our calculator uses Z for simplicity but notes the t-equivalent for small samples.
Worked example: A sports scientist measures VO2 max for 40 athletes. Sample mean = 52 mL/kg/min, standard deviation = 8 mL/kg/min, 95% confidence level.
Interpretation: We are 95% confident the true mean VO2 max for this athlete population is between 49.52 and 54.48 mL/kg/min.
How margin of error changes with sample size and confidence level (standard deviation = 10, mean = 50):
| Sample Size (n) | 90% CI Half-Width | 95% CI Half-Width | 99% CI Half-Width |
|---|---|---|---|
| 10 | ±5.20 | ±6.20 | ±8.14 |
| 25 | ±3.29 | ±3.92 | ±5.15 |
| 50 | ±2.33 | ±2.77 | ±3.64 |
| 100 | ±1.64 | ±1.96 | ±2.58 |
| 200 | ±1.16 | ±1.39 | ±1.82 |
| 500 | ±0.74 | ±0.88 | ±1.15 |
| 1,000 | ±0.52 | ±0.62 | ±0.81 |
Notice: doubling the sample size doesn't halve the margin of error — it reduces it by a factor of √2 ≈ 1.41. To halve the margin of error, you need to quadruple the sample size. This diminishing return is why very large samples provide only incremental precision gains.
Confidence intervals are frequently misinterpreted, even in published research. Here are the most common errors:
WRONG: "There is a 95% chance the true mean lies within this interval."
RIGHT: "If we repeated this study many times, 95% of the resulting intervals would contain the true mean."
The distinction matters because the true population mean is a fixed (unknown) value — it either is or isn't in any particular interval. The 95% refers to the long-run behavior of the interval procedure, not the probability for any single interval.
Other common misinterpretations:
These three concepts are frequently confused in research reports and data presentations:
| Measure | What It Describes | Formula | When to Use |
|---|---|---|---|
| Standard Deviation (SD) | Spread of individual data points | s = √[Σ(xi−x̄)²/(n−1)] | Describing variability in raw data |
| Standard Error (SE) | Precision of the sample mean | SE = s / √n | Comparing means across studies |
| Confidence Interval | Range for the population mean | x̄ ± Z × SE | Making statistical inferences |
A classic mistake in scientific papers: reporting "mean ± SD" when "mean ± SE" or "95% CI" would be more appropriate for inferential claims. SD tells you about the spread of your sample; CI tells you about your uncertainty in the population mean. For presenting inferential statistics, CIs are almost always more appropriate than SD.
Example: 50 runners have mean finish time of 4:15 per km, SD = 0:30 min/km. SD tells you individual runners vary by about ±30 sec/km around the mean. SE = 0:30 / √50 = 0:04.2 — the mean estimate is precise to ±4 seconds. The 95% CI = 4:15 ± 0:08.3 = [4:06.7, 4:23.3] — the true mean pace for this population is between 4:07 and 4:23 per km.
Confidence intervals appear across science, business, medicine, and everyday measurement:
Clinical trials: A new drug reduces systolic blood pressure by 12 mmHg [95% CI: 9–15 mmHg] compared to placebo. The entire interval is above zero, confirming statistical significance. The clinical team also considers whether even the lower bound (9 mmHg) is clinically meaningful.
A/B testing: Website version B shows 3.8% conversion rate vs version A's 3.2%. The 95% CI for the difference is [0.1%, 1.1%], excluding zero — so the improvement is statistically significant. The business team uses the lower bound (0.1%) as a conservative revenue impact estimate.
Sports science: An elite marathon training program shows an average 8-minute improvement in marathon time [95% CI: 5–11 minutes]. A runner deciding whether to adopt this protocol knows the benefit is likely 5–11 minutes, not just the point estimate of 8.
Quality control: A manufacturing process produces bolts with mean diameter 10.2 mm [99% CI: 10.15–10.25 mm]. If the specification requires 10.0–10.5 mm, the entire interval is well within spec, suggesting the process is in control.
Economics: GDP growth rate estimated at 2.3% [90% CI: 1.8–2.8%]. The interval helps policymakers understand the range of likely outcomes for budget planning.
In all these applications, reporting a point estimate without a confidence interval gives a false sense of precision. The interval honestly communicates the inherent uncertainty in any sample-based estimate — which is the foundation of honest, reproducible science.
A 95% confidence interval means: if you repeated your study many times, drawing new random samples each time, 95% of the resulting intervals would contain the true population mean. It does NOT mean there's a 95% probability the true mean is in this specific interval — the true mean is fixed; only the intervals vary from sample to sample.
Use the formula: CI = x̄ ± 1.96 × (s / √n), where x̄ is your sample mean, s is the standard deviation, and n is the sample size. For example: mean = 75, SD = 12, n = 50 → SE = 12/√50 = 1.697 → margin = 1.96 × 1.697 = 3.33 → CI = [71.67, 78.33].
A CI gets wider with: higher confidence level (99% vs 95%), larger standard deviation, smaller sample size. A CI gets narrower with: lower confidence level (90% vs 95%), smaller standard deviation, larger sample size. The most effective way to narrow a CI is to increase sample size — though the benefit diminishes as a square root.
Use the t-distribution when: your sample size is small (n < 30), and/or you don't know the true population standard deviation (which is almost always the case). The t-distribution produces wider intervals for small samples, which is more conservative and appropriate. For n ≥ 30, Z and t give nearly identical results. Our calculator uses Z for all sample sizes as a practical approximation.
The margin of error is the half-width of the confidence interval — the ± part. If CI = [71.67, 78.33], the margin of error is ±3.33. The confidence interval is the full range: [mean − margin, mean + margin]. In surveys and polls, "margin of error" typically refers to the half-width, while academic reports usually present the full interval.