Number of subjects needed for a precise inter-rater reliability estimate
This calculator finds the number of subjects needed to estimate an intraclass correlation coefficient (ICC) — a one-way random-effects model with k raters/replicates per subject — with a desired confidence interval width, using Bonett’s (2002) precision-based sample size formula.
Required subjects across plausible expected ICCs and total CI widths, using the currently selected number of raters and confidence level.
| Expected ICC | CI width 0.20 | CI width 0.30 | CI width 0.40 |
|---|
Sample size was planned to estimate a one-way random-effects intraclass correlation coefficient (ICC) using Bonett’s precision formula, assuming the specified expected ICC, number of raters/replicates per subject, two-sided confidence level, and total confidence-interval width. The reported sample size is rounded up to a whole number of subjects; Bonett’s k = 2, ICC ≥ 0.70 correction is applied where applicable. Citation: Bonett DG. Sample size requirements for estimating intraclass correlations with desired precision. Statistics in Medicine. 2002;21:1331–1335.