R help - Mar 2009 - ICC question: Interrater and intrarater variability (intraclass correlation coefficients)

Hello dear R help group.

I encountered this old thread (http://tinyurl.com/dklgsk) containing the a
similar question to the one I have, but left without an answer.
I am and hoping one of you might help.


A simplified situation: I have a factorial design (with 2^3 experiment
combinations), for 167 subjects, each one has answered the same question
twice (out of a bunch of "types" of questions).
Each answer could get an integer number between 0 to 3.

I wish to combine the two answers, but first to be sure I could, I would
have liked to run an ICC (Intraclass correlation) check on the two answers.
Naturally, I would use the irr, condord, or psy packages (as John Fox
suggested back then), but I can't because of the repetitions of different
design question for each patient.
Since the mentioned packages (irr, condord, and psy) Can take only a n*m
matrix for subjects and raters. But no place is given for the repetitions as
data and therefore it will be impossible to get results
for the INTRArater reliability.


Thanks,

Tal









-- 
----------------------------------------------


My contact information:
Tal Galili
Phone number: 972-50-3373767
FaceBook: Tal Galili
My Blogs:
http://www.r-statistics.com/
http://www.talgalili.com
http://www.biostatistics.co.il

	[[alternative HTML version deleted]]

Hello dear R help group.

I encountered this old thread (http://tinyurl.com/dklgsk) containing the a
similar question to the one I have, but left without an answer.
I am and hoping one of you might help.


A simplified situation: I have a factorial design (with 2^3 experiment
combinations), for 167 subjects, each one has answered the same question
twice (out of a bunch of "types" of questions).
Each answer could get an integer number between 0 to 3.

I wish to combine the two answers, but first to be sure I could, I would
have liked to run an ICC (Intraclass correlation) check on the two answers.
Naturally, I would use the irr, condord, or psy packages (as John Fox
suggested back then), but I can't because of the repetitions of different
design question for each patient.
Since the mentioned packages (irr, condord, and psy) Can take only a n*m
matrix for subjects and raters. But no place is given for the repetitions as
data and therefore it will be impossible to get results
for the INTRArater reliability.


Thanks,

Tal

	[[alternative HTML version deleted]]