Return to the research question first
Statistics are not a contest to display the largest number of tests. Before reading a table, identify the question, variables, unit of analysis, comparison, and evidence required for a responsible conclusion.
Separate three levels of interpretation
| Level | Purpose | Examples |
|---|---|---|
| Descriptive | Summarize the observed sample | Mean, median, frequency, standard deviation |
| Inferential | Represent uncertainty beyond the sample | Confidence intervals or a suitable test |
| Practical | Explain magnitude and usefulness | Effect size, direction, and real-world relevance |
Statistical significance is not practical importance
A small p-value does not automatically mean a meaningful effect, especially with a large sample. A non-significant result may also reflect limited power or imprecise measurement. Report the direction, magnitude, uncertainty, and design limitations instead of turning one statistic into a sweeping claim.
Write the result as an argument
State the answer to the question, show the evidence that supports it, and explain the boundary of the claim. Avoid “proves” when the design is descriptive or cross-sectional, and disclose missing data or multiple comparisons when relevant.
Make tables readable
Give every table a useful title, define abbreviations, use consistent decimals, and point readers to the main pattern. Do not repeat every cell in prose. A clear table allows readers to audit the logic rather than merely trust the conclusion.
Frequently asked questions
Is a p-value enough to report a finding?
No. Interpret it with the design, sample size, effect magnitude, uncertainty, and substantive meaning.
May I remove a result that does not support the hypothesis?
No. Report it transparently and discuss plausible explanations and limitations.
References
- Association of American Public Opinion Research — Best Practices for Survey Research
- Boynton & Greenhalgh — Selecting, designing, and developing your questionnaire
- Rutgers — Developing Effective Questionnaires
- TEQSA — Academic and research integrity
- European Commission — Responsible use of generative AI in research