Date: Wed, 15 Sep 2010 15:37:12 -0400
Reply-To: Art@DrKendall.org
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: Art Kendall <Art@DrKendall.org>
Organization: Social Research Consultants
Subject: Re: Generalized Estimating Equations (Clustering)
In-Reply-To: <AANLkTi=2jewWniZ1XvyHVOBaL9ndaCvxk3YQ-y8Hu_-H@mail.gmail.com>
Content-type: text/html; charset=ISO-8859-1
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
I read the OP a little differently. I believe the should be an
additional IV nZygotes to distinguish dizyotic and monozygotic
twins.<br>
Art<br>
<br>
On 9/15/2010 10:38 AM, R B wrote:
<blockquote
cite="mid:AANLkTi=2jewWniZ1XvyHVOBaL9ndaCvxk3YQ-y8Hu_-H@mail.gmail.com"
type="cite">
<div>Simon,</div>
<div> </div>
<div>I do not have time to read your entire post carefully, but I
think I have read enough to provide some [hopefully] useful
feedback to help you get started. Suppose your data set is
structured as:</div>
<div> </div>
<div>ID Twin X1 Y </div>
<div>1 1 24 0 </div>
<div>1 2 36 1</div>
<div>2 1 16 1 </div>
<div>2 2 14 1</div>
<div>3 1 22 0 </div>
<div>3 2 10 1</div>
<div>.</div>
<div>.</div>
<div>.</div>
<div> </div>
<div>ID = Dyad Identification Number</div>
<div>Twin = Twin Indicator</div>
<div>X1 = Continuous Predictor</div>
<div>Y = Binary Dependent Variable</div>
<div> </div>
<div>If you wanted to test for the effect of X1 on the binary
dependent variable, Y, while accounting for correlation of
residuals obtained from Twins, then you could fit a generalized
linear model using the following code:</div>
<div><br>
GENLIN Y (REFERENCE=FIRST) WITH X1<br>
/MODEL X1 INTERCEPT=YES<br>
DISTRIBUTION=BINOMIAL LINK=LOGIT<br>
/REPEATED SUBJECT=ID WITHINSUBJECT=Twin SORT=YES
CORRTYPE=EXCHANGEABLE ADJUSTCORR=YES COVB=ROBUST <br>
/PRINT CPS DESCRIPTIVES MODELINFO FIT SUMMARY SOLUTION.</div>
<div> </div>
<div>Note that ID is the SUBJECT variable, Twin is
the WITHINSUBJECT variable, and the specified correlation type
is EXCHANGEABLE. The EXCHANGEABLE type assumes there is residual
corrrelation, while the default INDEPENDENT type does not assume
any such correlation. </div>
<div> </div>
<div>If your dependent variable is continuous, then I suggest you
consider fitting a linear mixed model via the MIXED procedure. I
prefer not to comment any further in this particular post.</div>
<div> </div>
<div>HTH,</div>
<div> </div>
<div>Ryan<br>
<br>
<br>
</div>
<div class="gmail_quote">On Tue, Sep 14, 2010 at 11:44 PM, Simon -
<a moz-do-not-send="true" href="mailto:slmartys@gmail.com"
target="_blank">slmartys@gmail.com</a> <span dir="ltr"><<a
moz-do-not-send="true" href="mailto:slmartys@gmail.com"
target="_blank">slmartys@gmail.com</a>></span> wrote:<br>
<blockquote style="border-left: 1px solid rgb(204, 204, 204);
margin: 0px 0px 0px 0.8ex; padding-left: 1ex;"
class="gmail_quote">* Brief review of my project: I'm
investing MZ & DZ twin pairs and using<br>
two dichotomous categorical variables to examine differences
on several<br>
different IVs (some categorical). My analytic strategy is to
use SPSS GEE<br>
to account for non-independence of twin pairs.<br>
<br>
* Question 1: What is the appropriate "working correlation
matrix"?<br>
<br>
I have a variable that identifies each individual as belonging
to one dyad.<br>
I am using this variable as the "Subject" variable on the GEE
"Repeated"<br>
tab. I have another variable that arbitrarily designates one
twin as Twin 1<br>
and one as Twin 2. I have added this variable to
"Within-Subjects." What<br>
should I specify as the working correlation matrix? Someone
advised me that<br>
"robust estimator" is appropriate for covariance matrix, as
well as the<br>
"independent working correlation matrix", but I am not certain
that this is<br>
correct. Obviously, the "within-subjects" variable is
arbitrarily<br>
designating one half of the sample as "1" and one half as "2".
Does the<br>
independent working correlation matrix account for this
arbitrary<br>
designation? (In simple terms, what assumptions does it
make?)<br>
<br>
* Question 2: Would I need to consider changing the nature of
the working<br>
correlation matrix depending on the type and distribution of
my IV?<br>
<br>
* Question 3: The distribution of several of my IVs is
extremely, extremely<br>
skewed and is comparable to the Poisson distribution. Would
selecting the<br>
Poisson log be more robust than log-transforming the IV and
using a linear<br>
distribution? (I realize this is a very general question; I
just wondered<br>
how careful I need to be.)<br>
<br>
=====================<br>
To manage your subscription to SPSSX-L, send a message to<br>
<a moz-do-not-send="true"
href="mailto:LISTSERV@LISTSERV.UGA.EDU" target="_blank">LISTSERV@LISTSERV.UGA.EDU</a>
(not to SPSSX-L), with no body text except the<br>
command. To leave the list, send the command<br>
SIGNOFF SPSSX-L<br>
For a list of commands to manage subscriptions, send the
command<br>
INFO REFCARD<br>
</blockquote>
</div>
<br>
</blockquote>
</body>
</html>
=====================
To manage your subscription to SPSSX-L, send a message to
LISTSERV@LISTSERV.UGA.EDU (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
| 
