A significance test asks whether a sample result would be unusual if a specified population claim were true. For a single proportion, the null hypothesis supplies the reference value and the standard error used to measure the sample's departure from that value.
State hypotheses about the population
Let be the population proportion with the characteristic of interest. A test begins with and an alternative such as or
The hypotheses concern , not the observed sample proportion Choose the direction from the research question before examining which result appears most favorable. The null includes equality because it specifies the model used to calculate the reference distribution.
Check the test conditions
Use data from a random sample or a design supporting the intended inference. Check independence, including the 10% condition for sampling without replacement when appropriate. For a normal approximation under the null, require and
Unlike a confidence interval, this large-count check uses the null proportion The test asks what happens if the null is true, so its reference variability comes from that null model.
The one-proportion statistic is
It measures how many null standard errors the sample estimate lies above or below the hypothesized proportion.
Example: test for an increase
A service historically has a 40% renewal rate. In a random sample of 200 current customers, 94 renew. Test against
The sample estimate is 0.47. Under the null, expected counts are 80 renewals and 120 nonrenewals, so the large-count condition holds. The standard error is giving
The right-tail P-value is approximately 0.0217. At significance level reject the null. The data provide statistically significant evidence that the population renewal rate exceeds 40%, assuming the sampling and independence conditions are justified.
At the same evidence would not meet the rejection threshold. The data and P-value do not change when the chosen decision threshold changes.
A standard normal density with the region to the right of z = 2.02 shaded.
Under the null model, the shaded right tail contains z values at least as large as the observed 2.02. Its area is the P-value.
Describe this figure
A standard normal density with the region to the right of shaded.
Lines: observed z.
Interpret the P-value conditionally
The P-value is the probability, assuming the null model, of obtaining a statistic at least as extreme as the observed one in the direction of the alternative. In the example, it is the chance of a sample proportion at least 0.47, approximately, if the true proportion is 0.40.
It is not the probability that the null is true, not the probability the result happened “by chance” in a vague sense, and not the probability a future sample will repeat the result.
For a two-sided alternative, count departures in both directions. With a symmetric normal reference distribution, the two-sided P-value is twice the tail beyond A one-sided P-value should not be doubled unless the alternative is two-sided.
Example: failing to reject is not proving equality
Suppose a different sample gives for a two-sided test. Its P-value is about 0.549, so at the 5% level you fail to reject the null. The sample does not provide convincing evidence of a difference.
Do not conclude that the population proportion is exactly A small sample or a modest true effect can produce inconclusive evidence. “Fail to reject” describes the strength of the evidence, not a proof of the null claim.
Statistical significance is not practical importance
A very large sample can detect a tiny difference that has little practical value. Report the observed effect size, such as a 7-percentage-point increase in the example, alongside the test conclusion. A confidence interval can communicate plausible effect sizes and uncertainty.
Try it yourself
A two-sided test produces Approximate the P-value and state the decision at
Answer: The lower-tail probability is about 0.0062, so the two-sided P-value is about 0.0124. Reject the null at 5%. The negative statistic indicates the observed sample proportion is below the null value, even though the alternative allowed either direction.
What to remember
Specify the population parameter, use null-based variability, match the tail to the alternative, and interpret evidence without turning a P-value into a probability that a hypothesis is true.