12-21-2015 09:32 PM
We have a stratied sampling study. We can calculate the sample Intraclass Correlation Coefficient (ICC) from a two-way repeated-measures analysis of variance (using PROC MIXED) . Can we calculate the ICC at population-level using PROC SURVEYREG which strata design and the sampling weight are considered, that is, how to get the between-subject and total variance from PRC SURVEYREG? Or is it appropriate to using the MIXED model adding WEIGHT statement (applying the samplying weight) to get the ICC?
Your help is greatly appreciated!