|Date: ||Mon, 1 Dec 2003 11:58:44 -0700|
|Reply-To: ||"Schaible-Brandon,Sonia" <email@example.com>|
|Sender: ||"SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>|
|From: ||"Schaible-Brandon,Sonia" <firstname.lastname@example.org>|
|Subject: ||Re: intraclass correlations and correcting standard errors|
This is the article that made this concept make sense to me..
From: Nico Peruzzi, Ph.D. [mailto:email@example.com]
Sent: Monday, December 01, 2003 11:31 AM
Subject: intraclass correlations and correcting standard errors
I'm helping a colleague with a review he just got back from
a journal submission. In a nutshell, 19 doctors in each of
intervention and control groups saw about 3,000 patients
(each group saw 3,000), and they measured cost of drugs
prescribed. A reviewer noted that he should conduct an
Intraclass correlation to account for similarities among
patients nested within docs, and if that ICC is 0.10 or
more to adjust the results.
I've found some decent references that I'm going to pull to
get the theory behind the idea, but I hoped that some
listers might have crossed this bridge before and might
save me some time chasing down dead ends.
I see that I can pull the ICC from the Reliability menu,
and I found one reference recommending use of the One-Way
Random Model. However, if an adjustment is then required,
I don't yet have a clue as to how to proceed. Note that
the data was originally analyzed using Mann Whitney U for
continuous data and Chi Squares for categorical data.
Any tips or leads would be greatly appreciated.
Do you Yahoo!?
Free Pop-Up Blocker - Get it now