Rich Ulrich wrote:
>
> On Sat, 09 Dec 2000 00:02:06 +0800, haroldng <[EMAIL PROTECTED]>
> wrote:
>
> > Hi,
> > Could anyone tell me how should I interpret a negative ICC, say, e.g.,
> > ICC is 0.52 (95%CI: -0.15, 0.73).
>
> When you calculate a negative ICC as a point estimate,
> it indicates that you have observed more variation in a "class"
> than you expected by chance;
> the F-test between Subjects is less than 1.0.
> (You might get this if your sample had been "stratified"
> and your ANOVA failed to account for that.)
>
> The interpretation of a negative ICC in the range of the CI is
> "not statistically significant." Your sample is so small or your
> reliability is so poor, you haven't beaten chance. tsk-tsk.
Also, there are situations when you might expect more variability than due to
chance and the negative ICC is quite important (e.g., in some situations
clusters are systematically more dissimilar than one would expect by chance -
perhaps because units in the cluster compete for finite resources).
Thom
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
