What does a 95% confidence interval mean?

If you repeated the sampling process many times, about 95% of the calculated intervals would contain the true population mean. It does NOT mean there's a 95% probability the true value is in this specific interval — that's a Bayesian interpretation. The interval either does or doesn't contain the truth; we just don't know which.

How does sample size affect the confidence interval?

Larger sample sizes produce narrower confidence intervals — more precise estimates. Doubling sample size reduces margin by √2 (about 30%). Quadrupling n cuts margin in half. So getting 4× the data only gives 2× the precision; diminishing returns.

What confidence level should I use?

95% is the standard in research and reporting. 90% for exploratory analyses. 99% for high-stakes decisions (medical, safety). 99.9% for engineering or safety-critical applications. Higher confidence = wider intervals = less precise but more cautious.

When should I use t-distribution vs z?

Use t-distribution for small samples (n 30) or when population SD is known. As sample size grows, t approaches z. The t-distribution has heavier tails to account for additional uncertainty.

Why is my CI so wide?

Possible reasons: small sample size, high variability in data, or high confidence level. To narrow CI: collect more data (4× the data = half the margin), use lower confidence level (95% vs 99%), or reduce measurement variability. Wide CI signals less precise estimate.

How do I compare two CIs?

Not by overlap alone. Use proper t-test or compute CI for the difference. CIs may overlap but the difference between means could still be significant. CIs that don't overlap suggest a real difference (but rigorous test still needed).

Confidence Interval Calculator

Q: What's the difference between CI and prediction interval?

Confidence interval estimates the population parameter (mean, proportion). Prediction interval estimates the range for an individual future observation. Prediction intervals are wider than CIs because they account for both uncertainty about the parameter AND uncertainty about individual variation.

Enter the sample mean, standard deviation, sample size, and confidence level to calculate the confidence interval. Shows the margin of error and the upper and lower bounds.

A confidence interval (CI) is a range of plausible values for an unknown population parameter, computed from a sample. Unlike a single point estimate (sample mean), a CI captures the uncertainty inherent in estimating from a sample. The 95% confidence interval is the most common — it's the range that would contain the true population value 95% of the time across repeated sampling.

This calculator returns the confidence interval for a population mean given the sample mean, standard deviation, sample size, and desired confidence level. Margin of error = z × (SD / √n). Lower bound = mean − margin; upper bound = mean + margin. A 95% CI uses z = 1.96 for the multiplier.

Common confusion: the 95% does NOT mean "95% probability the true value is in this interval" (that's a Bayesian statement). It means: if we repeated the sampling process many times, 95% of computed intervals would contain the true value. The interval itself either does or doesn't contain the true value — we just don't know which.

Confidence intervals are the practical alternative to p-values: they show both the magnitude of the effect and the precision of the estimate. A narrow CI = precise estimate; wide CI = uncertain estimate. Both convey more information than significance tests alone.

Inputs

Sample Mean

Standard Deviation

Sample Size (n)

Confidence Level

Results

Lower Bound

46.4215

Upper Bound

53.5785

Margin of Error

± 3.5785

Standard Error

1.8257

Z-Value

1.960

Last updated: May 29, 2026

Formula

**Confidence interval for population mean:** CI = x̄ ± (z* × SD / √n) Where: - **x̄**: sample mean - **z***: critical z value for confidence level - **SD**: sample standard deviation - **n**: sample size **Margin of error:** ME = z* × SD / √n **Critical z-values:** | Confidence | z* | |---|---| | 90% | 1.645 | | 95% | 1.960 | | 99% | 2.576 | | 99.9% | 3.291 | **Worked example: Sample mean 50, SD 10, n = 30, 95% CI** ME = 1.96 × 10 / √30 = 1.96 × 1.826 = 3.58 Lower: 50 - 3.58 = 46.42 Upper: 50 + 3.58 = 53.58 95% CI: (46.42, 53.58) **For small samples (n < 30): use t-distribution instead** CI = x̄ ± (t* × SD / √n) Where t* depends on degrees of freedom (n - 1). **Sample size effect:** - Doubling n reduces margin by ~30% (factor of √2). - Quadrupling n cuts margin in half. - Tripling n by 1/3 reduces margin by ~13%. **Width vs sample size table:** | n | Margin (relative) | |---|---| | 25 | 1.00× | | 100 | 0.50× | | 400 | 0.25× | | 1000 | 0.16× | **Interpretation:** "We are 95% confident the true population mean is between 46.4 and 53.6." This means: - Across repeated sampling, 95% of intervals would contain the true value. - NOT: there is a 95% probability the true value is in this specific interval. - The interval either does or doesn't contain the true value. **Choosing confidence level:** | Level | Z* | Use case | |---|---|---| | 90% | 1.645 | Rough estimates, exploratory | | 95% | 1.960 | Standard in research | | 99% | 2.576 | High-stakes decisions | | 99.9% | 3.291 | Engineering, safety critical | Higher confidence → wider interval → less precise but more cautious. **Common applications:** - **Sample surveys**: estimating population proportion (±X% margin of error). - **Clinical trials**: estimating treatment effect. - **Quality control**: estimating manufacturing tolerances. - **Polling**: candidate support estimates. - **Scientific measurements**: physical constants with uncertainty. **Confidence interval vs hypothesis test:** - **CI**: gives range of plausible values plus magnitude. - **Significance test**: yes/no decision about specific value. - **Equivalence**: if 95% CI excludes null value, p < 0.05. - **CI is preferred**: more information than significance alone. **Calculation steps:** 1. Calculate sample mean (x̄). 2. Calculate sample SD. 3. Calculate standard error: SD / √n. 4. Look up critical z (or t for small samples). 5. Calculate margin: z* × SE. 6. Calculate bounds: x̄ ± margin. **One-sided confidence intervals:** Sometimes you only care about one direction: - **Upper bound**: x̄ + z* × SE (with z* for upper tail only) - **Lower bound**: x̄ - z* × SE Use one-sided when you have prior reason for direction. **Common sources of error:** - **Wrong critical value**: 1.96 for 95%, not 1.645 (90%). - **Population vs sample SD**: use sample SD in most cases. - **Wrong distribution**: t for small samples, z for large. - **Skipped finite population correction**: for large samples from small population. **Software:** - **Excel**: =CONFIDENCE.NORM(α, SD, n) for half-width. - **R**: t.test()$conf.int. - **Python**: scipy.stats.t.interval(). - **SPSS**: descriptive statistics options.

How to use this calculator

Enter your sample mean.
Enter sample standard deviation.
Enter sample size (n).
Select desired confidence level (90%, 95%, or 99%).
Calculator returns confidence interval and margin of error.
For n < 30, this gives approximation; use t-distribution for precise.

Worked examples

Survey response analysis

**Scenario:** Survey of 100 customers shows mean satisfaction 7.5/10. SD: 1.5. Calculate 95% CI. **Calculation:** ME = 1.96 × 1.5 / √100 = 1.96 × 0.15 = 0.29. CI: (7.5 - 0.29, 7.5 + 0.29) = (7.21, 7.79). **Result:** 95% CI: 7.21 to 7.79. We are 95% confident the true mean customer satisfaction in the population is between 7.21 and 7.79. Narrow CI indicates precise estimate with this sample size.

Clinical drug trial

**Scenario:** New drug tested on 60 patients. Mean reduction in symptoms: 25%, SD: 12%. 99% confidence interval needed for FDA submission. **Calculation:** ME = 2.576 × 12 / √60 = 2.576 × 1.549 = 3.99. CI: (25 - 3.99, 25 + 3.99) = (21.01, 28.99). **Result:** 99% CI: 21.0% to 29.0%. We are 99% confident the true symptom reduction is between 21% and 29%. FDA submission notes both the effect (25% reduction) and the precision (4% margin at 99% confidence).

Manufacturing process check

**Scenario:** Sample of 36 parts measures average weight 100.2 g with SD 1.5 g. Calculate 95% CI. **Calculation:** ME = 1.96 × 1.5 / √36 = 1.96 × 0.25 = 0.49. CI: (100.2 - 0.49, 100.2 + 0.49) = (99.71, 100.69). **Result:** 95% CI: 99.71 to 100.69. Manufacturing average is precisely measured (narrow CI of 0.98g). If spec is 100 ± 2, process is well-centered and consistent.

When to use this calculator

**Use confidence intervals for:**

- **Estimating population parameters**: mean, proportion, difference between groups. - **Reporting research findings**: alongside (or instead of) p-values. - **Polling and surveys**: margin of error in poll results. - **Quality control**: process capability estimates. - **Drug trials**: efficacy and safety estimates. - **A/B testing**: comparing conversion rates.

**CI interpretation rules:**

1. **CI captures uncertainty**: wider = more uncertain. 2. **95% means**: across many samples, 95% would contain true value. 3. **Either it does or doesn't**: this specific interval may or may not contain the truth. 4. **Doesn't mean "95% chance"**: the parameter is fixed; only intervals vary.

**Relation to hypothesis testing:**

- **CI excludes 0**: corresponds to p < 0.05 for difference test. - **CI excludes null value**: significant at chosen α level. - **CI overlaps**: not necessarily means no difference (but suggestive).

**Reporting:**

Always include: - Sample mean (point estimate). - Confidence level used. - Confidence interval bounds. - Sample size.

Example: "Mean revenue per customer was $50.20 (95% CI: $46.42–$53.58, n = 100)."

**Sample size considerations:**

- **n = 30**: minimum for normal approximation; t-distribution otherwise. - **n = 100**: typical for surveys. - **n = 1000**: high precision for national polls. - Larger n → narrower CI (precision).

**Margin of error common levels:**

| ME (%) | Required n (for proportion) | |---|---| | 10% | ~100 | | 5% | ~400 | | 3% | ~1067 | | 2% | ~2401 | | 1% | ~9604 |

(For 95% confidence, p = 0.5; this is conservative.)

**Special cases:**

- **Population proportion**: use p (proportion) and √(p(1-p)/n). - **Difference of two means**: combine variances appropriately. - **Regression coefficients**: software calculates automatically. - **Categorical data**: chi-square distribution intervals.

**Common misinterpretations:**

❌ "There's a 95% probability the true mean is between 46.4 and 53.6." ✓ "The true mean is fixed; if we sampled many times, 95% of computed CIs would contain it."

❌ "The CI tells me how confident I am in my data." ✓ "The CI quantifies uncertainty in the estimate of the population parameter."

❌ "Wider CI means worse data." ✓ "Wider CI means less precise estimate; could be due to small sample, more variability, or higher confidence level."

**CI vs prediction interval:**

- **CI**: range for population parameter (mean, proportion). - **Prediction interval**: range for individual future observation. - Prediction interval is wider than CI (more uncertainty for individual values).

**Bootstrapping:**

For non-normal data or complex statistics: - **Bootstrap**: repeated resampling with replacement. - Calculate statistic from each resample. - 2.5th and 97.5th percentiles give 95% CI. - No distribution assumption required.

**Bayesian credible intervals:**

Bayesian framework gives: - **Credible interval**: range where true parameter lies with X% probability. - This IS a probability statement about the parameter. - Different interpretation but similar numerical results often.

**One-sample vs two-sample:**

- **One-sample CI**: estimate single population mean. - **Two-sample CI**: estimate difference between two means. - **Paired sample CI**: estimate difference for matched pairs.

Common mistakes to avoid

Stating "95% probability true value is in interval." Incorrect frequentist interpretation.
Using z for small samples. Use t-distribution for n < 30.
Forgetting to convert CI bounds to original units. Standard CIs are computed in z-units.
Comparing CIs from different sample sizes directly. Width varies with √n.
Treating non-overlap as significance test. Use proper t-test for comparison.
Reporting only point estimate. Always include CI for transparency.
Confusing CI with prediction interval. CI is for parameter; prediction is for new observation.

Frequently Asked Questions

Sources & further reading

SponsoredShop Top Deals on AmazonSupport CalcMountain — browse top-rated products at no extra cost to you.