Statistics for Researchers › Module 6: Reporting and reasoning
Reporting results right
Lesson 1 of 3 · 2 min
A good results sentence lets a reader judge the finding without guessing. It needs a few parts.
The parts
- The estimate: the difference or effect, in real units.
- Its uncertainty: a 95% confidence interval.
- The sample: how many, in each group.
- The test: which one, with the exact p-value (not just "p < 0.05").
- An effect size, when units aren't obvious.
Weak versus strong
| Weak | Strong |
|---|---|
| The method was significantly better (p < 0.05). | The method improved accuracy by 2.1 points (95% CI 0.8 to 3.4; 5 seeds per method; paired t-test, p = 0.007). |
| Scores were higher in group B. | Group B scored 6.2 points higher on average (n = 48 and 51; 95% CI 1.9 to 10.5; d = 0.45). |
- The method was significantly better (p < 0.05).
- Strong: The method improved accuracy by 2.1 points (95% CI 0.8 to 3.4; 5 seeds per method; paired t-test, p = 0.007).
- Scores were higher in group B.
- Strong: Group B scored 6.2 points higher on average (n = 48 and 51; 95% CI 1.9 to 10.5; d = 0.45).
Also report
- Everything you tested, including results that didn't work out.
- Exclusions: data you removed and why.
- Software and versions, so others can repeat it.
The APA style guide and many journals now ask for effect sizes and confidence intervals, not just p-values.
RememberEstimate, interval, sample size, test, exact p. If a reader has to guess any of these, the result is incomplete.