Prev: Free Project Management Software Trial Demo
Next: Loop
From: cat.. on 11 Jun 2010 02:32
Hi there,

I need to calculate the Intraclass correlation coefficient between
series of data measured on the same patients (ratings of a scale by
several raters - more than 2).

I found that paper "Forming inferences about ICC"

http://www3.uta.edu/faculty/ricard/COED/McGraw%20(1996)%20Forming%20inferences%20about%20ICCs.pdf

It shows several way to calculate the ICC, all based on ANOVA models
of type:

x(i,j) = m(i,j) + e(i,j)

where x (i,j) is the jth measure on patient i

m(i,j) is the equation of the model.
Eg m(i,j) = mu + p(i) + r(j) (p = patient effect, r = rater
effect, mu = overall mean)

e(i,j) is the error term.

My concern is that validity assumptions of the ANOVA model include
independance of residuals (error terms). Which is not the case here as
all the error terms related to a single patient cannot not be
correlated.

Who has experience in ICC estimation and understood that ?

Many thanks in advance.

Catherine.
 | 
Pages: 1
Prev: Free Project Management Software Trial Demo
Next: Loop