Skip to content
Omnikalc

Fail-safe N calculator (Rosenthal)

Enter the number of studies, the average z-score of the studies and the significance threshold to get the number of unpublished studies with a null result that would be needed to make the combined result non-significant.

Updated on
Your inputs

Value out of range

Value out of range

Results

Fail-safe N

223

Combined z (Stouffer)
7.275
Tolerance level 5k + 10
70
Fail-safe N over the tolerance level
3.18 ×
Robust: hidden null studies are unlikely to overturn the result
Robust (1) or fragile (0)
1

The fail-safe N calculator gives Rosenthal’s file-drawer number: how many unpublished null studies would be needed to make a combined result non-significant, with the 5k + 10 tolerance check.

How the calculation works

  • Enter the number of studies and their mean z-score.
  • Choose the significance threshold.
  • The calculator gives the fail-safe N, the combined z, the tolerance level and a robustness verdict.

Example

Twelve studies have a mean z of 2.1, so the sum of z is 25.2 and the combined z is 25.2 / √12 = 7.275. With a one-sided 5 % threshold of 1.645, N = (25.2 / 1.645)² − 12 = 223. The tolerance level is 5 × 12 + 10 = 70, so the result is robust, with a margin of 3.18 times.

Why it matters

The calculator helps students and reviewers who assess publication bias, researchers who write the limitations of a meta-analysis, and teachers of research methods.

Good to know

The fail-safe N has well-known limits: it assumes that the hidden studies have exactly zero effect, whereas missing studies are often small positive or negative ones, and it ignores study size. It also does not estimate how much bias there is. Use it as a rough first check, along with a funnel plot and a statistical test of asymmetry. You can get the z-score of a study from its p-value, using the one-sided normal quantile. When the studies report different effect sizes, convert all of them to z-scores first, and keep the direction of each effect. Report the fail-safe N with the number of studies, the threshold used and the method of combining the z-scores, so that readers can reproduce it.

The meta-analysis calculator pools estimates, the z-score calculator converts values to standard units, the statistical power calculator shows why small studies matter and the p-value calculator links z and p.

Frequently asked questions

What is the fail-safe N?

It is the number of unpublished studies with an average effect of zero that would have to exist to bring the combined result back to non-significance. A large number suggests that publication bias alone is unlikely to explain the finding.

How is it computed?

With k studies of mean z, the sum of z is k × z. Adding N studies with z = 0 gives a combined z of Σz / √(k + N). Setting it equal to the threshold gives N = (Σz / z threshold)² − k.

What is the 5k + 10 rule?

Rosenthal suggested that a result is robust if the fail-safe N exceeds 5 times the number of studies plus 10. It is a rule of thumb, and many modern authors prefer funnel plots and methods such as trim-and-fill or selection models.

Was this calculator helpful?

Thanks for your feedback!

Results are estimates provided for information purposes. They do not replace professional advice.