Re: st:Why fixed effect estimates are different using Stata and SPSS?

[Steven Samuels <sjhsamuels@earthlink.net>]

MANDY,
I'm not experienced with SPSS, and the SPSS manual is not much help.  
However, I think that the Stata command that will produce results  
most comparable to your SPSS commands is -xtmixed- with the -reml-  
option. You noted that in your original post.  There are many options  
for estimating standard errors, and these might not be comparable.

-xtreg- with -fe- (or -re-)  does not use the REML algorithm to  
compute estimates and therefore will not reproduce  the SPSS figures.  
The models are also different quite different, as the -fe- command.  
You might get similar, if not identical, results from -xtreg- with  
the -mle- option and SPSS MIXED with /METHOD = ML.

=Steve

On Jan 6, 2009, at 7:50 PM, Mandy fu wrote:

> Hi Mr. Samuels,
>
> Thanks for  helping out! I really apprecaite it.
> Sorry to bother you again . After adding the statement that I forgot,
> the SPSS result is still different from Stata. It'll be great if you
> could give me some suggestion how to revise the commands to get the
> same result as Stata.  Thanks a lot!
> Mandy
>
> On Tue, Jan 6, 2009 at 7:10 PM, Steven Samuels  
> <sjhsamuels@earthlink.net> wrote:
> > The two setups are not comparable.  In your SPSS syntax, you left  
> > out a
> > statement like:
> > /RANDOM = INTERCEPT | SUBJECT(id)
> >
> > -Steve
> > On Jan 6, 2009, at 7:01 PM, Mandy fu wrote:
I find the results I got from estimating a simple fixed effect model
using Stata are quite different from what using SPSS. I was quite
confused that since SPSS estimates fixed effect using linear mixed
model (LMM), while in Stata when estimating LMM, I should use
-xtmixed- command instead of --xtreg--command.


*
*   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/