## Calculating ICC in proc mixed with CS covariance structure

I am using proc mixed with a CS covariance parameter structure (Type=CS).  Please tell me how to use the output to calculate the Intraclass Correlation (ICC).  In the past I used Type=UN and the covariance parameter estimates would include the Intercept and Residual.  Now the covariance parameter estimates include the Variance, CS, and Residual.

1.What do these three covariance parameter estimates (Variance, CS and Residual) refer to?

2. How can I calculate the ICC using them?

THANKS!!

Emily

3 REPLIES 3

## Re: Calculating ICC in proc mixed with CS covariance structure

Check this: Examples: Mixed Procedure

It give detials of how to do this with example

Search within this example for keyward [intraclass correlation coefficient]

## Re: Calculating ICC in proc mixed with CS covariance structure

Thank you so much for the reply.  The link you sent details ICC calculations for CS covariance structures, but in the output does not have the three lines I mentioned:  Variance, CS, Residual

Are you familiar with any sources that detail calculating the ICC using these covariance parameters?

Alternatively, is one of the three not used for calculating the ICC?

Thank you again!!

Emily

## Re: Calculating ICC in proc mixed with CS covariance structure

With TYPE=CS:

You can calulate ICC manually from the values in "Covariance Parameter Estimates" table as

ICC = Common Covariance Estimate /  (Residual Variance Estimate + Common Covariance Estimate)

Or use the RCORR option and let SAS to report the intraclass correlation coefficient in the "Estimated R Correlation Matrix"

`` REPEATED / SUBJECT=normalsubject TYPE=CS rcorr;``

.....

Discussion stats
• 3 replies
• 10149 views
• 0 likes
• 2 in conversation