Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: mixed multilevels models with weights


From   Steven Samuels <sjhsamuels@earthlink.net>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: mixed multilevels models with weights
Date   Sun, 27 Apr 2008 11:40:00 -0400

-gllamm- will accommodate weights for multilevel data. But, as Maarten says, the issue is complicated. For example, the weights provided with survey data usually only apply to the lowest level of the data--in your case, the student level. See: http:// www.cpc.unc.edu/restools/data_analysis/ml_sampling_weights . In the middle of the page is a link to a PDF file about this issue. It has list of packages that will handle weights. Austin Nichols (http:// www.statalist.eu/statalist/archive/2007-11/msg00812.html) noted that HLM should have been included.

SPSS is not one of the listed packages. The WEIGHT command in the SPSS MIXED procedure specifies frequency weights, not survey weights. If you use survey weights in the WEIGHT statement, they will add to the population size, and SPSS will think that this is your sample size.


-Steven



On Apr 27, 2008, at 10:09 AM, Maarten buis wrote:


--- Orsetta.CAUSA@oecd.org wrote:
I am using multilevel data coming from a cross-country dataset.
(student, school, country level).

I would like to estimate a mixed model, that is, containing fixed and
random intercepts, as well as fixed and random coefficients, taking
into account the nested structure of the data. I found a fantastic
tool with the xtmixed command. Until I discovered it did not handle
weights.

Yet, it is very important that I take the student weights into
account. I am not familiar with sas and spss, but apparently this
type of estimation, using weights, is possible. in spss the procedure
is called "mixed".

This suggests that it is not incorrect to take weights into account.

Does anyone have any suggestion?

Does by any chance the authors of the ado-file would be willing to
incorporate this feature? I am quite convinced that since multilevel
models often come from survey data, with the latter being always
provided with weights, this feature would be useuful to a number of
statausers.

Obviously, I cannot use the svy tool since the commands allowed by it
in terms of estimation are quite limited and do not include mixed
models.
Weights in a multilevel model is a remarkable complex issue, so if you
had to go for other software and had to choose between SPSS and SAS,
than I would have much more faith in SAS. If I indulge in a bit of
Collegestationology (as in Kremlinonlogy) than I would guess that if it
can be done, than it will be implemented at some point in the future,
because StataCorp has invested greatly in both -xtmixed- and its
capabilities of dealing with complex survey designs.

-- Maarten

-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands

visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434

+31 20 5986715

http://home.fsw.vu.nl/m.buis/
-----------------------------------------


__________________________________________________________
Sent from Yahoo! Mail.
A Smarter Email http://uk.docs.yahoo.com/nowyoucan.html
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   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