I want to compute an intraclass correlation in a multilevel regression model with a binary dependent variable (i.e., using logit). When a continuous DV is used, the ICC is the proportion of variance in the DV that is accounted for by the clustering (person-level) factor. Is the computation of the ICC the same when a binary DV is used?
I saw another post that dealt with this issue for count data as a DV (i.e., poisson): https://stats.stackexchange.com/a/173857/152743. I am not sure if this formula generalizes when a binary DV is used. As in this example, I am using the glmer function in R.