ICC reliability interval estimation

Expected ICC:
Number of raters (k):
Confidence interval % (1 – α):
Desired width of the confidence interval (w):
Required sample size (n):

Approximate number of subjects required to obtain an exact confidence interval of the desired width for an intraclass correlation coefficient (ICC) in one-way and two-way ANOVA models. Power (1 – β) fixed at 0.80.

Can be used to estimate the required approximate sample size (n) for inter-rater or test-retest reliability studies. For example: for a reliability study with two raters (or two repeated measurements) and an expected ICC value of 0.8, 51 subjects are needed to obtain a desired width of 0.2 (i.e., the value of ICC is between 0.7 and 0.9) for the 95% confidence interval.

Note: for k = 2 and ICCs ≥ 0.7, the approximate sample size tends to slightly diverge from the correct sample size (nc). In these special cases the accuracy can be considerably improved by adding (5 * ICC) to the approximate sample size. In the example above: nc = 51 + (5 * 0.8) = 55.

Reference:
Bonett DG. Sample size requirements for estimating intraclass correlations with desired precision. Stat Med. 2002;21(9):1331-5.