engleski

Intraclass correlation coefficient for grouped data

The maximum likelihood estimator of intraclass correlation coefficient (ICC) has not yet been suggested nor used in reproducibility studies with grouped data. Using data simulated from one-way random-effects model, we show that its bias is negligible regardless of the choice of categories, in contrast to commonly used midpoint ICC and weighted kappa. Implementation is provided in R package iRepro.

reproducibility; intraclass correlation coefficient; grouped data

nije evidentirano