Intraclass correlations (ICCs) and confidence intervals (CIs)

By Lynne Millar posted 02-03-2010 19:46

  
Hi Can anyone please provide advise on ICCs and CIs. If your ICCs are very low and include zero as the lower bound, would it be more appropriate to analyze at an individual level? Also are there any good references in relation to this problem?
1 comment
2 views

Related Content

Permalink

Comments

Christopher Tong
02-28-2010 18:08

Low ICCs could be an artifact of low between-subject variability, which has nothing to do with reliability. It's an artifact of the sample of subjects.
This would certainly warrant more detailed analysis, but it shows the inadequacy of ICC as a measure of reliability.
See for instance, Bland & Altman (1990), Computers in Biology and Medicine, 20: 337-340.
Dunn (2004) Statistical Evaluation of Measurement Errors, 2d ed.