 |
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: GLLAMM and weights
Second thoughts: Rescaled weights are not what your original article
called for. They are designed to reduce bias in the estimated
coefficients and variance components under a sampling model, not the
observation model that you are studying. Therefore, -gllamm- with
rescaled weights may be the wrong approach. You might write to the
authors and ask what software they used.
-Steve
On Dec 13, 2008, at 9:52 PM, Steven Samuels wrote:
I haven't tried this in GLLAMM, so I'm not sure what goes wrong
when you add the weights. Check that your version of -gllamm- is up-
to-date. As requested in the Statalist FAQ, show exactly what you
typed and what Stata produced. It is good practice, as others have
pointed out, to start with the simplest model possible (e.g. no
covariates) and try progressively more complex models (e.g. add
treatment) to see where the problem starts. You probably have not
rescaled the weights as the GLLAMM manual suggests. See: http://
www.stata.com/meeting/4nasug/Chantala.ppt and http://
www.cpc.unc.edu/restools/data_analysis/ml_sampling_weights. These
contain links to the Stata program -pwigls- which will scale the
weights. Rabe-Hesketh and Skrondal (2006) compute the "Method 1"
weights by hand and illustrate an analysis in GLLAMM. Try to
replicate their results.
-Steve
Rabe-Hesketh, S. & Skrondal, A. (2006). Multilevel modelling of
complex survey data. Journal of the Royal Statistical Society:
Series A (Statistics in Society), 169(4), 805-827.
On Dec 13, 2008, at 2:06 PM, Ariel Linden wrote:
I posted a question a couple weeks ago about using multiple
weights per
person/period data (that is, using a different weight per person
for each
period). The problem was that xtgee will not accept different
weights per
person.One suggestion that was provided was to use GLLAMM. I tried
that and the
multiple weights were indeed accepted by the model. However,
another issue
presented itself. I am using 3 variables for the fixed effects;
wave, treatment (dummy 1/0),
and wave * treatment (for growth). Unfortunately, the model drops
both the
wave and wave * tx variables. This does not happen when I run an
unweighted model, nor when I run this
using regress or glm. Of course, I need both those variables in
the model.
*
* 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/