ICC from lme / reliability coefficient

[Karin <twentestat@yahoo.com>]

Hello,

 

My (statistical) question is on an estimate of the intraclass correlation in a repeated measures design. I want to use that correlation as an reliability coefficient.

 

The design is as follows:

10 persons are all repeatedly (10 times) measured under four different treatment/conditions combinations. And this whole experiment is done twice for all persons. So in theory we have 10*2*2*2 = 800 observations.

I use the function lme to analyse the data (and not raov function since I have some missing data).

Anyhow next to the effect of the treatment and condition I want to say something about the reliability of the measurement.

 
I use the following syntax:

fit_lme(fixed=resp~condition*treatment,random=~1|persons,data="">

I get an estimate of the stdev of the intercept and the residuals. Is it correct to use

the ratio of the variance of the intercept and the sum of the variances of the intercept and residuals as an estimate of the intraclass correlation coefficient?

 

Or am I doing something wrong?

 

Thanks for any comments,

 

Karin

---

Do you Yahoo!?
Yahoo! Mail Address AutoComplete - You start. We finish.