Power calculation for the likelihood ratio-test when comparing two dependent intraclass correlation coefficients

Comparing the reproducibility level of two devices with continuous outcome on a unique sample of subjects (each subject being assessed several times with both devices) comes down to compare two dependent intraclass correlation coefficients (ICCs). When planning such a reproducibility study, one has to specify both the number of subjects to be included and the number of replicates per subject associated to each device. We propose SAS and S-plus macros, which allow power calculations by implementing a simulation study where dependent ICCs are compared by means of a likelihood ratio-test.