Dear all,
I would like to compute ICCs to estimate the variance in various measures attributable to membership in the 20+ groups in which those measures were administered. However, I'm unclear about which ICC is most appropriate (two-way vs. one-way, mixed vs. random, absolute agreement vs. consistency). The logic described in the literature that I've read about choosing ICCs when raters are involved does not seem to apply here. I would greatly appreciate any thoughts and suggestions that you have. Thank you.