Writing Complete Conclusions for Mean Inference Problems

From a Test Decision to a Complete Conclusion

A mean test ends with more than “reject” or “fail to reject.” The final conclusion should answer the research question in context: what population mean or difference in means is being studied, what direction the evidence supports, and whether the data provide convincing evidence for the claim. In “Tying the Conclusion to the Original Claim” and “Common Mistakes in Writing Conclusions,” you practiced these ideas. Here, you will apply them across one-sample, paired, and two-sample settings.

The core wording stays consistent across these settings, but the parameter changes. A one-sample test concerns one population mean, \(\mu\). A paired test concerns the population mean of the pairwise differences, \(\mu_d\). A two-sample test concerns a difference between two population means, such as \(\mu_1-\mu_2\). Make that parameter clear in the conclusion rather than relying on a vague phrase such as “the average is different.”

A Reliable Shape for the Final Sentence

As in “Making a Decision From P-Value and Alpha,” reject \(H_0\) when \(p\le\alpha\), and fail to reject \(H_0\) when \(p>\alpha\). The evidence statement follows that decision: rejection supports saying the data provide convincing evidence for the alternative; failure to reject calls for saying they do not provide convincing evidence for the alternative. Neither decision proves a claim.

A useful structure is: “Since \(p\) [comparison] \(\alpha\), we [decision]. The data [do/do not] provide convincing evidence that [contextual alternative claim about the population parameter].” You may combine the two parts into one sentence, but both should be clear. Include the numerical p-value and \(\alpha\) when they are provided or requested.

The final point matters because statistical evidence and study design answer different questions. As explained in “Generalizing Conclusions to a Population,” sampling determines what population a result may represent. Random assignment can support a causal interpretation of treatment differences; random sampling can support generalization to a population. A p-value alone provides neither kind of support.

One-Sample Mean: Name the Population and Benchmark

For a one-sample test, the conclusion should identify the population whose mean is being tested and the benchmark in the alternative claim. Do not substitute the sample mean for the population mean. The sample mean is evidence used in the test; the conclusion concerns the population mean.

Paired Mean: Make the Difference’s Sign Meaningful

A paired test is a one-sample t test applied to pairwise differences. As in “Two-Sample Versus Paired Test on the Same Numbers,” the data’s genuine pairing determines the procedure. A conclusion should state what a positive or negative difference means in the setting. Define the subtraction order consistently with the test’s alternative; otherwise, a correct decision can be followed by a claim in the wrong direction.

If the subtraction order were reversed, the numerical differences and alternative direction would also need to be reversed. For example, defining \(d=\text{new-system time}-\text{usual-system time}\) would make evidence for a lower mean time with the new system correspond to \(\mu_d<0\), not \(\mu_d>0\). Always interpret the defined difference rather than memorizing “greater means better” or “less means worse.”

Two-Sample Mean: Keep the Group Order in the Conclusion

For two independent groups, define the difference in the same order used by the hypotheses. If the parameter is \(\mu_1-\mu_2\), a positive difference means group 1’s population mean is greater than group 2’s, and a negative difference means it is lower. The conclusion must preserve this order. The full test structure and parameter definitions were covered in “Writing the Full Two-Sample t Test Solution” and “Defining Both Population Means in Context.”

When the Evidence Is Not Convincing

A fail-to-reject conclusion is not a weaker version of a rejection conclusion. It answers the question carefully: the observed data do not provide convincing evidence for the alternative claim at the chosen significance level. It does not say that the null hypothesis is true, that the means are identical, or that an effect is impossible. As discussed in “Why a t Test Never Proves the Null Mean,” a test may fail to detect a difference even when one exists.

For example, suppose a paired test asks whether a new stretching routine reduces athletes’ mean recovery time, and \(p=0.12\) at \(\alpha=0.05\). The comparison is \(0.12>0.05\), so the decision is to fail to reject \(H_0\). A suitable conclusion would say that the data do not provide convincing evidence that the routine reduces the population mean recovery time for the athletes represented by the study. It would not say “the routine has no effect.”

The same evidence wording applies whether the alternative is one-sided or two-sided, but the claim must match that alternative. For a one-sided test, name the specified direction. For a two-sided test, say “differs” or “is different,” not “is higher” unless the test was actually designed to support that directional claim.

Common Mistakes and AP Exam Tips

• Stopping at the decision: “Reject \(H_0\)” does not answer the research question. Add a contextual statement about what the data provide convincing evidence for.
• Using the sample result as the conclusion: “The sample mean is lower” reports a statistic, not the population claim. A full-credit conclusion names the true population mean or mean difference.
• Reversing a paired difference: If \(d\) is defined as before minus after, interpret its sign using that order. State what a positive or negative difference means before writing the claim.
• Reversing two groups: For \(\mu_1-\mu_2\), keep group 1 first and group 2 second. “Group 2 is lower” may be mathematically equivalent to “group 1 is higher,” but unclear switching makes the direction hard to grade.
• Claiming equality after failing to reject: A large p-value is not evidence that two population means are exactly equal. Say the data do not provide convincing evidence for the alternative.
• Overstating what the design supports: Do not claim causation from an observational comparison, or generalize to a population not represented by the sampling method. Avoid “proves,” “always,” and “for everyone.”

A quick final check is to read your conclusion without looking at the hypotheses. Can a reader tell which population or groups are being discussed, what quantitative variable is measured, and whether the evidence supports a higher, lower, or different mean? If not, revise the contextual claim. Then check the study description: does it justify the population scope and any causal wording you used?