Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

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

st: -xtgee- with multiple clusters

From   Andreas Jensen <>
Subject   st: -xtgee- with multiple clusters
Date   Thu, 29 Jul 2010 23:21:08 +0200

Hi Statalist.

I'm working on a data set with a three-level cluster structure. In 12
countries (fixed effect) a number of households have been sampled
(level 3). In each household its members (level 2) have answered a
questionnaire on two occasions/waves (level 1). I do not have a
balanced design since some member of each cluster drop out and new
ones are added during the waves.

I have computed a mixed effect logistic model with one random
intercept over households (modeling the correlation of members inside
each household) and one random intercept over household members
(modeling the longitudinal correlation).

My problem is that the model takes far too long time to compute due to
the random effects and the sample size. Now I want to try with a
marginal model being aware that the interpretation of the coefficients
is different.

Looking at various examples of the -xtgee- command (e.g. in the Hardin
and Hilbe book) I see that they specify i(id) and t(t) options. I
reckon that this specifies an id for (in my case) people and waves and
the GEE approach will take care of the correlation in y_it for fixed i
and different values of t.

In my case I have i = householdID, j = memberID and t = waveID and the
following correlations seems plausible cor(y_ijt, y_ij't), cor(y_ijt,
y_ijt') and cor(y_ijt, y_ij't').

Any pointers on how to accomplish such a model in STAT 11 if it's possible?

Best regards,
Andreas Jensen
*   For searches and help try:

© Copyright 1996–2016 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index