TimelyScholarFree Tools › P-Curve Analysis
Enter the focal test from each study (for example t(28) = 2.9) to see the distribution of significant p-values and test whether the set of findings contains evidential value.
FreeNo account neededYour data stays in your browser
| Right-skew, full p-curve | Z = -0.85, p = 0.196 |
|---|---|
| Right-skew, half p-curve | Z = -0.70, p = 0.241 |
| Flatter than 33% power, full | Z = -0.91, p = 0.181 |
| Flatter than 33% power, half | Z = 2.35, p = 0.991 |
The p-curve is inconclusive: it shows neither clear right skew nor clear flatness.
Include only one statistically independent, focal result per study, and use the exact test statistics reported. Results match the published p-curve app (version 4.06).
Formats: t(df)=x, F(df1,df2)=x, r(df)=x, z=x or chi2(df)=x.
Compare the observed p-curve with the curves expected under no effect and under 33% power.
Copy the right-skew and flatness tests for the full and half p-curves.
P-curve looks only at statistically significant results (p < .05). If the studied effects are real, small p-values (such as p < .01) should be more common than p-values just under .05, giving a right-skewed curve. A flat or left-skewed curve suggests the findings lack evidential value or may reflect selective reporting.
The half p-curve uses only results with p < .025, which makes it more robust to ambitious p-hacking. Choose one focal test per study, decided before looking at the results, and document your choices in a disclosure table.
No. P-curve needs the test statistic and degrees of freedom so it can compute each result's probability under different effect sizes.
There is no fixed minimum, but the tests have little power with only a handful of significant results; interpret small p-curves cautiously.