How to use it
Enter the mean exactly as printed in the paper — trailing zeros matter.
A mean written 2.50 was rounded to two decimals; 2.5 to one,
and the test depends on knowing which.
Items averaged is 1 for a single integer measurement. If each participant's score is itself the mean of several integer items (a five-item Likert subscale, say), enter the number of items — the granularity becomes 1/(N × items), and ignoring this is the most common source of false positives.
What a failure means
It proves the reported numbers cannot all be correct. It does not indicate fraud. In practice the ranking is: a typo, a misreported sample size, unreported missing data, then — rarely — fabrication. See why an impossible number is usually a typo.
Read more
- The GRIM test, explained — the maths, a worked example, and the traps
- GRIMMER calculator — the same idea applied to standard deviations
- How to check the statistics in a paper