Thank you Philip for your comment.
As the aim of analysis is estimating the rate of change in BMI over time, a 3-level-model is employed. We would also like to look at the differences in the rate of change between cases and controls.
BMI data for each individual collected over time. However, as data are not a random sample of the population and come from a nested-matched-case-control study, we would like to take into consideration similarities between cases and matched controls as a level in the analysis.
Looking forward to hearing from other comments.
Ghasem
Philip McShane <[log in to unmask]> wrote:
Ghasem
Is this necessary or even appropriate? Presumably you are interested in
knowing whether BP differs between cases and controls, What's wrong with
a paired t-test?
Regards
Phil McShane
-----Original Message-----
From: A UK-based worldwide e-mail broadcast system mailing list
[mailto:[log in to unmask]] On Behalf Of ghasem yadegarfar
Sent: 01 May 2007 19:13
To: [log in to unmask]
Subject: 3-level model in SPSS with second level as fixed effect
Dear Allstat,
I am carrying out a 3-level repeated measures model in MlWiN and SPSS
as follows:
level 1: each BMI reading
level 2: group (cases or controls) data come from a cohort based
case-control study
level 3: each pair of matched cases and controls (about 1000 pairs)
In MlWiN it is very easy to keep level 2 (group) as a fixed effect.
However, I am struggling with SPSS and I could not get exactly the same
answer as I got from MlWiN.
Your ideas how to run a 3-level model in SPSS with second level as
fixed is very appreciated. Other comments, particularly on residual
covariance type, are most welcome.
All the best
Ghasem
---------------------------------
Yahoo! Mail is the world's favourite email. Don't settle for less, sign
up for your freeaccount today.
---------------------------------
Yahoo! Answers - Got a question? Someone out there knows the answer. Tryit now.
|