# How to Write a Confidence Statement (With Examples)

A confidence statement is a sentence that reports an interval estimate and the confidence level attached to it. The safest wording is "We are 95% confident that the true mean lies between X and Y," because the confidence belongs to the procedure that produced the interval, not to the parameter itself. This article shows you how to write a confidence statement step by step, with a worked example and the phrasing errors that reviewers flag most often.

## Quick Answer

- State the confidence level, the parameter, the units, and both bounds in one sentence.
- Use "We are 95% confident that..." or "The 95% confidence interval for the mean is (X, Y)."
- Never say "There is a 95% probability that the true mean is between X and Y." Once the interval is computed, the parameter is either inside it or not [1].
- Report the point estimate and the interval together, for example "mean 12.36 s, 95% CI (12.01, 12.71)."
- Round the bounds to the same precision as the estimate, and always give the units.

## Before You Start

You need four things before you can write the sentence.

First, the parameter you are estimating. It might be a population mean, a proportion, a difference between two means, or a regression coefficient. The statement changes with the parameter, so name it precisely.

Second, the confidence level. Most reports use 95%, but 90% and 99% are common too [2]. The level is a property of the procedure. If you sample the same population many times and build an interval each time, about 95% of those intervals would bracket the true parameter value [3].

Third, the interval itself, meaning the lower and upper bounds. The bounds are the numbers at the two ends of the interval [4]. If your mean is 12.36 with bounds 12.01 and 12.71, the interval is 12.01 to 12.71.

Fourth, the units and the context. "Between 12.01 and 12.71" means nothing without "seconds" and without saying whose reaction time you measured.

One more decision: one-sided or two-sided. A two-sided interval brackets the parameter from above and below. A one-sided interval gives only an upper or lower bound [3]. Most reports use two-sided intervals, so that is what the rest of this article assumes.

If you need to compute the interval first, the [Confidence Interval Calculator](/tools/confidence-interval-calculator) handles the arithmetic, and [What Is a Confidence Interval? Formula and Examples](/blog/data-analysis/confidence-interval-formula-examples) walks through the formula.

## Step by Step

1. **Identify the parameter and its symbol.** Write down whether you are estimating $\mu$, $p$, $\mu_1 - \mu_2$, or a coefficient $\beta$. The statement must name the same thing.

2. **Record the confidence level.** Write "95%" explicitly. Do not leave the level implied, because readers cannot guess whether you used 90%, 95%, or 99%.

3. **Write the point estimate with units.** For a mean, this is $\bar{x}$. For a proportion, it is $\hat{p}$. Include the unit, such as seconds, milligrams, or percentage points.

4. **Write the interval in parentheses.** Use the format (lower, upper) with a comma, or "lower to upper." Keep the same number of decimal places on both bounds.

5. **Choose a phrasing template.** The two standard forms are:
   - "We are 95% confident that the true mean reaction time lies between 12.01 and 12.71 seconds."
   - "The mean reaction time was 12.36 seconds (95% CI: 12.01 to 12.71)."

6. **Check the direction of the claim.** The interval describes the parameter, not the sample. The sample mean is a fixed number you computed. The interval is your estimate of where the population value sits.

7. **Add the interpretation only if the reader needs it.** In a technical paper, the numbers plus the confidence level are usually enough. In a report for a general audience, one plain sentence explaining that the interval reflects sampling uncertainty helps.

The general form of the interval is:

$$CI = \text{point estimate} \pm \text{margin of error}$$

where the margin of error is the critical value times the standard error of the estimate [5].

## Worked Example

The dataset is 25 lab trials recording reaction times in seconds.

| trial | reaction_time_s | trial | reaction_time_s | trial | reaction_time_s |
|---|---|---|---|---|---|
| 1 | 10.2 | 10 | 12.4 | 19 | 12.5 |
| 2 | 11.5 | 11 | 13.7 | 20 | 11.4 |
| 3 | 12.1 | 12 | 12.0 | 21 | 12.7 |
| 4 | 12.8 | 13 | 11.8 | 22 | 13.2 |
| 5 | 13.4 | 14 | 12.9 | 23 | 11.7 |
| 6 | 11.9 | 15 | 13.3 | 24 | 12.3 |
| 7 | 12.6 | 16 | 11.6 | 25 | 13.5 |
| 8 | 13.1 | 17 | 12.2 | | |
| 9 | 11.2 | 18 | 13.0 | | |

Here are the steps with the computed values.

| Step | Value |
|---|---|
| Sample size n | 25 |
| Sample mean x̄ | 309.00 / 25 = 12.3600 |
| Sample SD s (n-1) | 0.8377 |
| Standard error SE | 0.8377 / sqrt(25) = 0.1675 |
| t critical (df = 24, 95%) | 2.0639 |
| Margin of error | 2.0639 × 0.1675 = 0.3458 |
| Lower bound | 12.3600 - 0.3458 = 12.0142 |
| Upper bound | 12.3600 + 0.3458 = 12.7058 |

The interval is (12.01, 12.71) after rounding to two decimals.

The correct confidence statement is:

> We are 95% confident that the true mean reaction time lies between 12.01 and 12.71 seconds.

The wrong statement is:

> There is a 95% probability that the true mean reaction time lies between 12.01 and 12.71 seconds.

The second sentence is wrong because the true mean is a fixed number. Once you have computed the interval, the parameter is inside it or it is not. The 95% refers to the long-run success rate of the method, not to a probability about this one interval [1]. Confidence intervals and confidence levels are frequently misunderstood, and published studies have shown that even professional scientists often misinterpret them [1].

Here is the code that produced these numbers.

```python
import numpy as np
from scipy import stats
times = [10.2, 11.5, 12.1, 12.8, 13.4, 11.9, 12.6, 13.1, 11.2, 12.4, 13.7, 12.0, 11.8, 12.9, 13.3, 11.6, 12.2, 13.0, 12.5, 11.4, 12.7, 13.2, 11.7, 12.3, 13.5]
n = len(times)
mean = np.mean(times)
sd = np.std(times, ddof=1)
se = sd / np.sqrt(n)
tcrit = stats.t.ppf(0.975, df=n-1)
moe = tcrit * se
ci = (mean - moe, mean + moe)  # (12.01, 12.71)
print(tuple(round(float(x), 2) for x in ci))
```

Output:

```
(12.01, 12.71)
```

## Other Ways to Do It

The template above works for a mean. Other parameters need slightly different wording.

**Proportion.** "The proportion of respondents who preferred the new layout was 0.62 (95% CI: 0.55 to 0.69)."

**Difference between two means.** "The treatment group scored 4.2 points higher on average (95% CI: 1.1 to 7.3)." Here the interval is for the difference, so zero is the value of interest. If the interval excludes zero, the difference is statistically significant at the 0.05 level [6].

**Regression coefficient.** "Each additional year of experience was associated with a 0.8% increase in salary (95% CI: 0.3% to 1.3%)."

**Odds ratio.** "The odds ratio was 1.45 (95% CI: 1.10 to 1.91)." An interval that excludes 1.0 indicates a statistically significant association at the 0.05 level.

**One-sided interval.** "We are 95% confident that the true mean is at least 12.07 seconds." Use this only when you have a directional hypothesis before seeing the data.

You can also express the interval in terms of statistical significance. The 95% interval contains exactly the parameter values that would not be rejected by a two-sided test at the 0.05 level [1]. That equivalence is useful when you want to connect your interval to a hypothesis test. If you are still framing the research question, [How to Write a Hypothesis for a Research Proposal: Examples and Templates](/blog/research-skills/how-to-write-a-hypothesis-for-a-research-proposal-examples-and-templates) covers that step, and [Statistical Questions Examples: How to Write Them](/blog/guides/statistical-questions-examples-how-to-write-them) helps you phrase the question the interval answers.

## Troubleshooting

**The interval is very wide.** Width depends on the confidence level, the sample size, and the variability in the sample [5]. A wider interval means less precision. Report it honestly instead of switching to a lower confidence level after the fact.

**The interval includes zero for a difference.** That means zero is a plausible value for the difference, so you cannot rule out no effect at your chosen confidence level [6]. Say so plainly.

**The interval includes 1.0 for a ratio.** The same logic applies to odds ratios and risk ratios, where 1.0 is the null value.

**Your software reports a z interval but you expected t.** The z critical value applies when the population standard deviation is known. In practice it usually is not, so you estimate it from the data and use t [3].

**The bounds look inconsistent with the estimate.** For a t interval around a mean, the estimate should sit exactly in the middle of a two-sided interval. If it does not, you may have mixed a one-sided interval with a two-sided estimate. Intervals for odds ratios, risk ratios and proportions are often asymmetric, so this check does not apply to them.

## Common Mistakes

- **Saying "95% probability" about the parameter.** The parameter is fixed. The probability statement applies to the procedure, not to this interval [1]. Fix: write "We are 95% confident that..." instead.
- **Omitting the confidence level.** A bare interval (12.01, 12.71) tells the reader nothing about precision. Fix: always attach the level, as in "95% CI."
- **Omitting units.** "Between 12.01 and 12.71" is ambiguous. Fix: add "seconds," "milligrams," or whatever the measurement is.
- **Reporting the interval without the point estimate.** Readers want to see the center as well as the range. Fix: report both, for example "12.36 s (95% CI: 12.01 to 12.71)."
- **Claiming the interval covers 95% of the data.** A confidence interval for the mean describes the mean, not the spread of individual observations. Fix: use a prediction interval or a reference range if you want to describe individual values.
- **Changing the confidence level after seeing the result.** Switching from 95% to 90% to make an interval exclude a null value is a form of p-hacking. Fix: set the level before analysis and report it as planned.

## Limitations

A confidence interval quantifies sampling uncertainty only. It does not correct for bias from a bad sampling frame, measurement error, confounding, or missing data. A narrow interval from a biased sample is still biased, and it can look more convincing than it deserves.

The interpretation also depends on the model behind the calculation. The interval for a mean assumes the sampling distribution of the mean is approximately normal, which usually holds for large samples by the central limit theorem but can fail for small samples from heavily skewed data. For proportions near 0 or 1, normal-approximation intervals behave poorly, and exact or adjusted methods are better [4]. Finally, the confidence level is a property of repeated sampling. It says nothing about whether your particular study was well designed.

## Frequently Asked Questions

### What is the correct way to write a confidence statement?

Name the confidence level, the parameter, the estimate, the interval, and the units. A complete version reads: "We are 95% confident that the true mean reaction time lies between 12.01 and 12.71 seconds." A compact version for a results table reads: "12.36 s (95% CI: 12.01 to 12.71)."

### Can I say "there is a 95% chance the true value is in this interval"?

No, not under the frequentist interpretation. Once the interval is computed, the true parameter is either inside it or outside it. The 95% describes how often the method produces intervals that capture the parameter across repeated samples [1]. A Bayesian credible interval does support a probability statement about the parameter, but that is a different construction.

### Should I report a 95% or 99% confidence interval?

Report 95% unless you have a reason to do otherwise. It is the most common level in biomedical and social science research [5]. Use 99% when you need stronger evidence against a null value and can accept a wider interval, or 90% when you want a narrower interval and can accept weaker evidence.

### How many decimal places should the interval have?

Match the precision of your point estimate. If you report the mean as 12.36, report the bounds as 12.01 and 12.71. Do not report four decimals for a measurement recorded to one decimal place. Consistent rounding keeps the statement readable and avoids implying precision you do not have.

### Does a confidence interval replace a hypothesis test?

It can. A two-sided 95% interval contains exactly the parameter values that a two-sided test at the 0.05 level would fail to reject [6]. Reporting the interval gives readers both the effect size and its precision, which is more informative than a p-value alone. Many journals now expect intervals alongside or instead of significance tests [1].

## References

1. [Confidence interval - Wikipedia](https://en.wikipedia.org/wiki/Confidence_interval)
2. [Inference: Confidence Intervals - ENV710 Statistics Review Website](https://sites.nicholas.duke.edu/statsreview/671-2/)
3. [7.1.4. What are confidence intervals?](https://www.itl.nist.gov/div898/handbook/prc/section1/prc14.htm)
4. [3.4: Confidence Limits - Statistics LibreTexts](https://stats.libretexts.org/Bookshelves/Applied_Statistics/Biological_Statistics_(McDonald)/03%3A_Descriptive_Statistics/3.04%3A_Confidence_Limits)
5. [Hazra A. (2017). Using the confidence interval confidently. Journal of thoracic disease](https://pmc.ncbi.nlm.nih.gov/articles/PMC5723800/)
6. [7.1.5. What is the relationship between a test and a confidence interval?](https://www.itl.nist.gov/div898/handbook/prc/section1/prc15.htm)

## Further Reading

- [Greenland S, Senn SJ, Rothman KJ et al. (2016). Statistical tests, P values, confidence intervals, and power: a guide to misinterpretations. European Journal of Epidemiology](https://doi.org/10.1007/s10654-016-0149-3)

## Related Articles

- [How to Calculate Confidence Level: Formula and Examples](/blog/data-analysis/how-to-calculate-confidence-level)
- [What Is a Confidence Interval? Formula and Examples](/blog/data-analysis/confidence-interval-formula-examples)
- [Statistical Questions Examples: How to Write Them](/blog/guides/statistical-questions-examples-how-to-write-them)
- [How to Write a Hypothesis for a Research Proposal: Examples and Templates](/blog/research-skills/how-to-write-a-hypothesis-for-a-research-proposal-examples-and-templates)