[R] GmeanRel vs. mult.icc - differential meaning of icc2 results?

[Alain D.]

Dear R-list,

let us say I want to report the reliability of group means for "HRS" (ICC2):

library(multilevel)

data("bh1996")
mult.icc(bh1996[,c("HRS","LEAD","COHES")],grpid=bh1996$GRP)

Variable     ICC1    ICC2
1      HRS 0.129237 0.91713
2     LEAD 0.147462 0.92804
3    COHES 0.048049 0.79008

the result is 0.91713. 

However, if I use GmeanRel I get a different result:

tmod<-lme(HRS~1,random=~1|GRP,data=bh1996)
mean((GmeanRel(tmod))$MeanRel)
[1] 0.89586

What is the difference between both estimates?

Thanks for helping me!

Best wishes

Alain 
 
	[[alternative HTML version deleted]]