Confidence Interval to Standard Error
Recover a normal-approximation SE on a difference or log-ratio scale from compatible interval limits.
Your entries stay in this page’s memory and are not submitted to us. Downloads are generated locally. Reloading or leaving the page clears your entries. Do not enter patient information.
Recover an approximate standard error only when a reported interval was constructed using a normal critical value and symmetry on the chosen analysis scale.
Calculation results
Approximate SE: 1.00002
Implied arithmetic interval midpoint: 3. Compare this with the reported estimate; disagreement can indicate rounding or an unsuitable conversion.
Method and limits
SE = (upper − lower)/(2z). For a ratio, use [ln(upper) − ln(lower)]/(2z), yielding the SE of the log effect, not the raw ratio. z is 1.644854, 1.959964, or 2.575829 for 90%, 95%, or 99% intervals.
This converter does not recover SD, sample size, or participant-level data. It is unsuitable for small-sample t intervals, exact intervals, bootstrap intervals, asymmetric profile-likelihood intervals, or Bayesian credible intervals. Check the original method before using a recovered SE in synthesis.
Cochrane: effect measures and calculations · Read the practical guide · Download blank worksheet (TXT)