1.1 KiB
1.1 KiB
Statistical Reporting Standard
Minimum reporting package
For every major comparison, report:
- metric definition and direction,
- unit of analysis,
- sample size / run count,
- descriptive statistics,
- uncertainty estimate,
- inferential test,
- effect size,
- correction strategy when multiple contrasts exist,
- limitation if assumptions or sample size are weak.
Required fields
Descriptive
mean ± stdwhen repeated runs are comparable95% CIwhen inference is discussed- median / IQR when distribution is strongly non-normal
Inferential
- exact test name
- test statistic and degrees of freedom when applicable
- p-value format
- effect size
- correction method for multiple comparisons
Do not do these
- report only best run
- report only p-values
- hide non-significant comparisons
- treat unstable trends as conclusions
- switch tests without stating why
Default wording rule
Use three layers:
- Observation — what changed numerically
- Support — what the test/effect size says
- Boundary — what remains uncertain