Calculate 95% CI for ICCs with SAS macro

[Shoriuchi]

Dear All

How can I calculate 95% confidence intervals for ICCs with SAS macro (%intracc) (http://support.sas.com/kb/25/031.html).

I am trying to calculate ICC(2,k) but %intracc but the macro produce only ICCs but not 95% CIs.

Thank you.

Best regards,

[RW9]

The text in red tells you what you need to know.  This macro is written by a non-SAS person.  You would need to contact them about specifics of the macro, or if you have the code, alter it yourself.  Yes, so you have the code:

http://support.sas.com/kb/25/addl/fusion25031_1_intracc.sas.txt

You need to go through that and see if the procedures involved can generate the numbers you want, and if so modify it to your needs.

[FreelanceReinh]

Hello @Shoriuchi,

I found a hint in one of my books that CIs were described in the paper Shrout and Fleiss (1979), which your linked Sample 25031 refers to. I googled for it (search terms: Shrout Fleiss 1979 "Intraclass correlations") and the very first link in the results led to a PDF scan of the article. It looks like combining a simple formula on p. 426 and complicated formulas on p. 424 might yield the CI you're looking for. So, if you're familiar with the parameters used in those formulas, you should be able to compute the confidence limits. The relevant F quantiles can be calculated using the FINV or QUANTILE function, as you probably know.

[Adele3]

Hello @FreelanceReinh,

Your response to this post has been very helpful as I am trying to calculate the 95% CI for the Shrout Fleiss ICC (3,1).

When calculating the relevant F quantiles using the FINV function, it is not clear what to input as the numerator degrees of freedom and denominator degrees of freedom. Are you able to provide clarity? 

[FreelanceReinh]

Hello @Adele3,

Let me answer your question in your thread PROC MIXED: ICC 95% CI, not in someone else's which should have been closed years ago.