Check my statistics: free online statcheck, p-value checker and GRIM test calculator (no sign-up)
Paste your Results section and every p is recomputed from its t, F, chi-square, r or z and its degrees of freedom, with the GRIM test on means and a cross-check of the N. Runs in your browser, free, no sign-up, nothing is uploaded.
Free · runs in your browser · no sign-up
Paste your Results section, or the whole manuscript, and every p is recomputed from its test statistic and degrees of freedom (t, F, χ², r, z). It is the statcheck check (Nuijten et al., 2016), plus the GRIM test on means of integer variables (Brown & Heathers, 2017) and a cross-check of the total N between sections. Nothing leaves your browser.
The text is processed in your browser and is never sent to a server.
What does not add up
Changes the conclusion: The reported p says one thing about significance and the recomputed p says the opposite: this changes what is concluded.
Inconsistent: The reported p does not follow from that statistic with those degrees of freedom. Usually a transcription slip, but a reviewer will see it.
GRIM: that mean is impossible with that n
The total N changes between sections
Chi-square without its N: Without the N in parentheses neither the effect size nor the table can be recovered. APA 7 asks for it: χ²(df, N = …).
p = .000 does not exist: A p is never exactly zero: it is a rounding of something below .0005. APA 7 asks for p < .001.
GRIM test by hand
If all you have is a mean and an n: a mean of integer data (ages in whole years, counts, single unaveraged items) with n participants can only take values that are multiples of 1/n.
With n > 100 or more than two decimals the test loses power and almost anything is possible. Scale means (the average of k items) admit multiples of 1/(k·n) and should not be checked this way.
What this does not see
It only recognizes statistics written in the text in APA form ("t(28) = 2.10, p = .045"). Those living in a Word table, in a figure or without their degrees of freedom never get here.
It does not recompute effect sizes or confidence intervals. To pull d, g, r or η² out of a sentence there is the effect size from text converter, and the check that the effect size matches the F is a separate one.
When the text does not allow a decision (one tail versus two, a p already corrected for multiple comparisons, df that could be the N) the entry is listed as not checkable rather than as an error. A false accusation is worse than silence.
GRIM is applied only to means the sentence itself declares integer (age in whole years, counts, cigarettes, sessions) with the n nearby and n ≤ 100. Scale means are not checked: almost any value is possible.
It does not read the design, whether the analysis answers the hypothesis, or the references. That is what a reviewer does, not a formula.
Each check, explained
Each one has its own page, with the formula, a real case and the standalone tool:
A program by Nuijten, Hartgerink, van Assen, Epskamp and Wicherts (2016) that recomputes the p of every APA-formatted statistic and compares it with the one in the text. Over 250,000 results in psychology, they found that half the articles had at least one inconsistency and one in eight had one that changed the conclusion. This page runs that check in your browser, with a more cautious rule: if the text does not allow a decision, it does not accuse.
What is the GRIM test?
Granularity-Related Inconsistency of Means (Brown & Heathers, 2017). With n participants and integer data, the mean is an integer sum divided by n, so it can only take certain values. A mean that is none of them is impossible with that n.
Is my text stored or sent anywhere?
No. The whole check runs in your browser; there is no call to any server and no account is needed. Close the tab and nothing remains.
What if I want someone to read the whole manuscript?
This page checks numbers. Previum reviews the whole manuscript: two reviewers, references checked against the databases and these same checks anchored to each sentence. The Basic review is free.
This checks the numbers. Previum reviews the manuscript.
Two reviewers, references checked against the databases and these same checks, anchored to the sentence that triggers them. The Basic review is free; the Professional review develops everything.
Brown, N. J. L., & Heathers, J. A. J. (2017). The GRIM test: A simple technique detects numerous anomalies in the reporting of results in psychology. Social Psychological and Personality Science, 8(4), 363–369. https://doi.org/10.1177/1948550616673876
Nuijten, M. B., Hartgerink, C. H. J., van Assen, M. A. L. M., Epskamp, S., & Wicherts, J. M. (2016). The prevalence of statistical reporting errors in psychology (1985–2013). Behavior Research Methods, 48(4), 1205–1226. https://doi.org/10.3758/s13428-015-0664-2