A confidence interval estimates an unknown population proportion while acknowledging random sampling variation. It combines a sample estimate with a margin of error. The interval's interpretation concerns the population parameter and the long-run reliability of the method.
Define the parameter and estimate
Let be the true proportion of the target population with a specified characteristic. From a sample of individuals with successes, the estimate is
A common one-proportion interval is
The square-root term is the estimated standard error, describing how much sample proportions vary. The critical value determines how many standard errors the interval extends on either side. For 95% confidence,
This method is an approximation. Its validity depends on the sampling design and enough successes and failures for the normal approximation.
Check conditions before calculating
The data should come from a random sample or another design supporting the intended inference. Observations should be independent; for sampling without replacement, check that the sample is no more than about 10% of the population.
For the usual large-count condition, require at least 10 observed successes and 10 observed failures: and Here the interval uses the sample estimate because the true is unknown.
A large sample cannot correct undercoverage, voluntary response, or misleading questions. The margin of error describes random sampling uncertainty under the model, not every possible source of error.
Example: construct and interpret an interval
A random sample of 400 residents includes 232 who support a proposal. Then with 232 successes and 168 failures. Assuming the independence condition is reasonable, the estimated standard error is
The 95% margin of error is The interval is approximately
A suitable conclusion is: “We are 95% confident that between about 53.2% and 62.8% of all residents in the target population support the proposal.” State the population, characteristic, and percentage range. Do not describe the interval as containing 95% of individual residents or 95% of sample responses.
What confidence means
If the same random-sampling method were repeated many times and a 95% interval constructed each time, about 95% of those intervals would contain the fixed population proportion. The intervals vary; does not.
After this particular interval is calculated, it either contains or it does not. The confidence level describes the method's long-run coverage, not a frequentist probability assigned to the fixed parameter being randomly inside these fixed endpoints.
Example: plan a sample size
For desired margin of error , solve
where is a planning estimate. If none is available, use 0.5 because is largest there, giving a conservative sample size.
For 95% confidence and margin at most 0.03, Round up to 1,068 completed responses. Rounding down would miss the stated guarantee. Expected nonresponse may require contacting more people than the required completed sample.
Understand the tradeoffs
Higher confidence requires a larger critical value and a wider interval for the same data. A larger sample produces a narrower interval. To halve the margin of error while keeping other factors fixed, multiply the sample size by four.
The confidence level and margin of error are different concepts. A 95% confidence interval can be narrow or wide depending on sample size and estimated variability.
Try it yourself
A sample yields with Find the approximate 95% margin of error, and explain what happens if becomes 400 with the same estimate.
Answer: The standard error is 0.05, so the margin is At equals 400, the standard error is 0.025 and the margin is 0.049, exactly half as large.
What to remember
Check the design and counts, compute estimate plus or minus margin, and interpret the interval for the population proportion. Confidence describes repeated-method coverage, while sample size controls precision.