Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: RE: Panel data - XTGEE with 2 levels of clustering


From   "lmaltadealencar" <lmaltadealencar@ucsd.edu>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: Panel data - XTGEE with 2 levels of clustering
Date   Thu, 2 Apr 2009 11:55:05 -0700

Dear Paul and Clyde,

Paul, the visits' order really won't matter (not a time series) but I do
need to consider the eyes, as they're the actual unit of analysis here. The
analysis is for an ophthalmology study - and eyes with more advanced disease
are expected to have a poorer performance on these tests, even when
comparing the best and the worst eye of a same patient, but the correlation
between these results is still higher than from one patient to another. So I
can't really go for a GLM.

Clyde, I did consider -xtmixed- instead of GEE, but opted for GEE only
because I'm really interested in the association of the two main variables,
and not on the random effects. Now I feel that's the best option and I'll go
on and use -xtmixed-.

Thank you both very much for the suggestions.  

Luciana




-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Clyde Schechter
Sent: Thursday, April 02, 2009 8:11 AM
To: statalist@hsphsun2.harvard.edu
Subject: st: Panel data - XTGEE with 2 levels of clustering

I don't think you can do this with -xtgee-.  You need -xtmixed-.

Without knowing more about the substance of your problem, I can't tell
exactly how you would specify the model.  But it looks like you have
visits nested within eyes nested within subjects.  After converting "eye"
to a numeric variable (-encode-) you will probably end up needing
something like:

xtmixed slabpeak wdtpeak || subj: || eye: || visit:

Clyde Schechter MA MD
Albert Einstein College of Medicine
Bronx, NY, USA

// BEGIN ORIGINAL MESSAGE
Date: Wed, 1 Apr 2009 17:07:35 -0700
From: "Malta De Alencar, Luciana" <lmaltadealencar@ucsd.edu>
Subject: st: Panel data - XTGEE with 2 levels of clustering

Dear Statalist,

I would like to ask some help on a multilevel unbalanced model (subjects -
eyes -
visits).
I have several subjects (subj). For most subjects both eyes were included
(eye), and
each eye was assessed from 1 to 3 different times (visits) under different
medications (the order of the visits is not important). Intraocular
pressure (IOP -
continuous variable) was measured by two different methods in each occasion
(iop_slab and iop_wdt).
I am interested in the comparison between these two methods (more
specifically if
wdt_iop can "predict" slab_iop), but I need to adjust for the repeated
measures in
each eye and for the correlation between both eyes of a same patient.

I?ve found some examples and publications online, but mostly with only two
levels. I
am relatively new to Stata. I?m using Stata/MP 10 on Windows. I believe
the best
option is to use xtgee, I'm only not sure how to include both levels.

I've pasted a sample of my data below:

subj eye visit wdtpeak slabpeak
22016 OD 1 25.50 18.50
22016 OS 1 19.50 19.50
22019 OD 1 16.50 22.50
22019 OS 1 15.25 21.00
22020 OD 1 23.00 22.00
22020 OD 2 20.00 20.00
22020 OS 1 23.50 23.00
22020 OS 2 18.00 20.50
22021 OD 1 33.00 34.50
22021 OD 2 31.00 30.50
22021 OS 1 26.50 30.00
22021 OS 2 25.00 27.50
22022 OD 1 18.00 23.00
22022 OD 2 16.00 23.00
22022 OS 1 20.00 25.50
22022 OS 2 17.50 23.00
22023 OD 1 21.50 22.50
22023 OD 2 21.50 23.50
22023 OS 1 20.00 22.00
22023 OS 2 20.00 23.00
22024 OD 1 22.00 22.00
22024 OD 2 19.00 22.50
22024 OD 3 18.00 23.00

Thank you very much.

Luciana



Luciana M. Alencar, M.D.
University of California , San Diego
Hamilton Glaucoma Center room 174
9500 Gilman Drive, Dept. 0946
La Jolla, CA 92093-0946
phone: 858.534.5334
fax: 858.822.0615
email: lalencar@glaucoma.ucsd.edu<mailto:atafreshi@glaucoma.ucsd.edu>

// END ORIGINAL MESSAGE

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index