Re: st: RE: Comparing Chi2/L2 in different samples using bootstrap

[Steven Samuels <sjsamuels@gmail.com>]

Since your question has now changed to one about multinomial logistic  
models, I suggest that you repost with a new subject heading.

Steve



Steve, Stas, thanks a lot for your time and input

I hear your points. Indeed, pooling the dataset sounds like a good  
option, but here's where the complicaton is: the model has six  
categorical variables and six continious ones. Interacting year with  
each predictor (and each category in case of the categorical vars.)  
will make the model quite complicated. Besides, I am not sure if I  
will be able to use the mlogview command in this case, which makes  
interpretation so much easier.
As for the clustering, the cluster size is actually quite big in each  
survey year - I think at least 40 observations. There's also at least  
50 clusters in each year in total. As pointed, I do svyset the data  
before. However, I am willing to sacrifice clustering option in the  
model
I haven't tried bsweights yet, but I will read on this later today.

After all, and in a nutshell if things are kept as they are, can I  
stretch my conclusions this far to compare the coefficients?

Thanks again for your help,
Dmitriy

________________________________________
From: Dmitriy Poznyak
Sent: Monday, December 06, 2010 2:44 PM
To: statalist@hsphsun2.harvard.edu
Subject: Comparing Chi2/L2 in different samples using bootstrap

Hello all,

I am estimating three identical multinomial models with bootstrap for  
the different years of survey data, for instance. 1991, 1999 and 2007.  
Aside from comparing predicted probabilities, which I assume shouldn't  
pose any problem, I need to compare Chi2/L2 coefficients for the  
different variables in the model. The rationale for doing this, is  
that the fit of the individual predictors (e.g. social-demographic  
stuff) declines through time. Here's where the question arises.  
Clearly, samples in different years have different size, and perhaps  
different design effects, and so on.

In order to possibly address these issues I ran the bootstrapped  
models with the same number of iterations in each case:
bootstrap, reps(2000) force: mlogit vote5  x y z ... ,base(1) cl(zip),  
[pweight=weight1], rrr
Next, I test the effect of the predictors:  test x;  test z, etc.  
Again, the models' specification is identical for all years; what  
differs is the sample size and design.

Considering the bootstrap method being used, will it be possible to  
compare Chi2/L2 and perhaps pseudo R2 coefficients for different  
samples in this case, and, if not, what would be my strategy. Note  
that pooling datasets is not feasible due to several reasons, like  
weighting, etc.

Thanks for your suggestions,
Dmitriy

*
*   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/