Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


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

Re: st: weighting ordinal multilevel model


From   Lucas <lucaselastic@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: weighting ordinal multilevel model
Date   Tue, 11 Feb 2014 07:36:54 -0800

Whether ignoring the sample design will give you wrong results depends
on: 1)the sample design and 2)your research question.

A low-level example is that if a sample of engineers oversamples women
in engineering, and if one's aim is compare male and female engineers,
then one need not use weights.  However, if one seeks to use that
sample to ascertain the average outcome for engineers, then one would
need to use weights.

As one multilevel model example of an infinite number of
possibilities, if the complexity of the sample design is such that
within-context the cases are a simple random sample of all cases in
the context, but contexts are sampled using stratified sampling, then
controlling for the context-level strata at level 2 should be
sufficient to avoid having to use weights.

I'm no sampling statistician, so if one is listening I invite them to
weigh-in (no pun intended).

For more examples with regression (which will suggest the principles
underlying my response above) see:

Winship, Christopher, and Larry Radbill.  1994.  "Sampling Weights and
Regression Analysis." *Sociological Methods and Research* 23: 230-257.

Take care.
Sam

On Tue, Feb 11, 2014 at 7:18 AM, Stas Kolenikov <skolenik@gmail.com> wrote:
> As is clear from -help meologit-, it does not support weights. If you
> are really interested in weighted estimation (and you should be),
> you'd have to use -gllamm-. As a devil's advocate, I don't know what
> the reference to Snijders & Boskers 2012 is, and what these guys
> discuss and suggest, so you would want to elaborate on that. Ignoring
> the sampling design will almost definitely give you wrong results.
>
>
> -- Stas Kolenikov, PhD, PStat (ASA, SSC)
> -- Principal Survey Scientist, Abt SRBI
> -- Opinions stated in this email are mine only, and do not reflect the
> position of my employer
> -- http://stas.kolenikov.name
>
>
>
> On Tue, Feb 11, 2014 at 10:06 AM, Pritsch, Julian <j.pritsch@mpicc.de> wrote:
>> Dear statalist,
>>
>> I want to estimate an ordinal multilevel model using Stata 13 and the meologit command. Is it possible to use this command with weighted data? My data result out of a complex survey design. I would prefer (following the weighting discussion in Snijders & Boskers 2012) not to use weights but at least I want to compare the results.
>>
>> Any advice on that matter if weighting in multilevel data is a good idea per se?
>>
>>
>> Cheers,
>> Julian
>>
>>
>>
>>
>> *
>> *   For searches and help try:
>> *   http://www.stata.com/help.cgi?search
>> *   http://www.stata.com/support/faqs/resources/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/faqs/resources/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/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index