Re: st: xtmixed output

[David Airey <david.airey@Vanderbilt.Edu>]

This is covered in Rabe-Hesketh and Skrondal's book available from  
Stata Press.

Note -xtreg- gives a z statistic as does -xtmixed- and -gllamm-. It is  
-regress- that give the t statistic.

According the authors, "The reason this statistic is called z instead  
of t is that a standard normal sampling distribution is assumed under  
the null hypothesis that B = 0 instead of a t distribution. The t  
distribution is a finite sample distribution whose shape depends on  
the degrees of freedom. For the variance components model, the finite  
sample distribution does not have a simple form, so Stata's commands  
used the asymptotic (large-sample) sampling distribution. (Some other  
software packages approximate the finite-sample distribution by a t  
distribution where the degrees of freedom are some function of the  
data.) ..."

and they go one from there...


On Oct 27, 2008, at 5:00 AM, Gabriele Schino wrote:

> Dear Statalisters,
>
> I have a (possibly silly) question about the output provided by  
> xtmixed.
>
> In testing the significance of the individual independent variables  
> xtmixed provides a z value and a P value. This is in contrast with,  
> for example, xtreg, that provides t, df and P values.
>
> My question is: given that my sample is obviously finite, shouldn't  
> xtmixed test significance using a statistic (like t) for finite  
> samples? How do I attach a df to my test?
>
> Thank you very much for your help.
>
> All the best,
> Gabriele Schino
>
> *
> *   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/