How to calculate ICC1 and ICC2 with unbalanced data?

[Victor Smith]

Hi all, I have a problem with the -icc- command. It omits a large number of observations with unbalanced data.

Is there a way to calculate ICC1 and ICC2 with unbalanced data in stata (includes all observations, with k being the average number of judges in each group)? ICC1 is the "individual" part in the results of -icc-, ICC2 is the "average" part in the results of -icc-.

[Joseph Coveney]

I'm not sure about individual versus average, but for unbalanced data, you can use a combination of mixed , reml dfmethod(kroger) and estat icc. (With judges and targets both random, you might have to compute ICC manually.)

[daniel klein]

Also, see kappaetc (SSC). It does not support average ICC but it handles unbalanced data (and repeated measurements).

Best
Daniel