On Tue, 26 Mar 2002 21:49:08 +0000 (UTC), "JP"
<Janepeutrell@removethis bit.btinternet.com> wrote:

> I have been unable to find an adequate explanation of exactly what is the
> difference between an interclass and in intraclass correlation, and the
> circumstances in which you would choose either.

For the 'intraclass correlation', you *assume*  that the means 
have the same mu.  It can be computed for a *set*  of raters.
Deviations are computed around the common mean.

Thus, you only use it for 2 ratings when you expect equality,
with parallel measurements.  The intra-class is what is possible,
when you can't  systematically  "distinguish"  the raters in
a fair way. 

For contrasting 2 raters, I like using a paired t-test and the
corresponding interclass correlation.  That shows you both
the main pieces of information, without confusing them or
confounding them at all.  You get r  to measure parallelism; 
you get t  to measure  mean-difference. 

(What you don't have assured  is the equal-variances, 
which presumably isn't much of a problem.)

-- 
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to