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

Re: st: GLLAMM using weights

From   Stas Kolenikov <>
Subject   Re: st: GLLAMM using weights
Date   Mon, 2 Nov 2009 15:17:46 -0500

For -gllamm-, you actually provide the prefix rather than the full
name with the -weight- option. Hence, you should have as many weight
variables as you have levels (if your lowest level is the actual
response, rather than a variable in a latent variable system with many
dependent variables). Ideally, you should have the weights for all the
levels, but if you don't (most likely), define the level 1 weights as
the sampling weights from CPS, and all other weights, as 1:

gen wgt1 = <name of CPS weight variable>
gen wgt2 = 1
gllamm ... , ... weight( wgt )

For some models, it is the relative ratios of the weights that
matters. Some refs:,,

On 11/2/09, Doug Hess <> wrote:
> I wish to use the weights provided in the Current Population Survey's
>  supplements to analyze binary responses by individuals (and sometimes
>  by households); I wish to use -gllamm- because the units are clustered
>  in states and I wish to test the impact of policies that vary by
>  states. However if I use the -weight(VAR)- command following the
>  -gllamm- commands, it says the weight variables cannot be found.
>  I guess I am naively using the -weight- command and the weight
>  variables just as I would if the command was -iweight- when analyzing
>  the data in -logit- without the clustering. Is this the wrong way to
>  think of this? Any solutions? (The CPS weights are ten digits long
>  including four decimal places).
>  Thanks for any leads.
>  Doug Hess
>  PhD Candidate, Public Policy
>  George Washington University
>  *
>  *   For searches and help try:
>  *
>  *
>  *

Stas Kolenikov, also found at
Small print: I use this email account for mailing lists only.
*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index