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

# RE: st: RE: Accessing random coefficients in xtmixed

 From "Martin Weiss" To Subject RE: st: RE: Accessing random coefficients in xtmixed Date Sat, 12 Dec 2009 10:10:43 +0100

```<>

Given all the confusion that we have observed over time regarding the coefficients returned by -xtmixed- and its siblings, the -coeflegend- option added in Stata 11 is of course highly appreciated. Unfortunately, in the -xtmixed- case, it does not extend to the random intercepts and slopes which would be a welcome addition...

*******
webuse pig, clear
xtmixed weight week || id: week, coefl
*******

HTH
Martin

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Sven-Oliver Spieß
Sent: Samstag, 12. Dezember 2009 00:32
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: RE: Accessing random coefficients in xtmixed

Hi Andrew,

It is true that Stata does not list the standard deviations in the
matrices. However it saves the logarithms of them in the e(b) matrix.
Thus you can easily get the desired estimates by exponentiating the
values in the matrices:
xtmixed depvar indepvar || groupvar: , cov(unstr) mle
di exp([lns1_1_1]_b[_cons])
di exp([lnsig_e]_b[_cons])

Beware that the order of the standard deviation of the random intercept
changes when you add random slopes.
Thus in a model with a random slope you'd have to access the SD of the
RI by typing:
xtmixed depvar indepvar || groupvar: indepvar, cov(unstr) mle
di exp([lns1_1_2]_b[_cons])	// --> RI now lns1_1_*2*

This behavior makes sense considering the convention that Stata always
lists constants in models last but at first this threw me slightly off.

Cheers,
Sven-Oliver

On Fri, 2009-12-11 at 23:28 +0100, Martin Weiss wrote:
> <>
>
> You have to undo Stata`s transformations, that much is true:
>
>
> *******
> webuse pig, clear
>
> xtmixed weight week || id: week
>
> nlcom(sd_week: exp(_b[lns1_1_1:_cons])) /*
>  */  (sd_cons: exp(_b[lns1_1_2:_cons])) /*
>  */  (sd_res: exp(_b[lnsig_e:_cons]))
> *******
>
>
> HTH
> Martin
>
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Andrew Spiegelman
> Sent: Freitag, 11. Dezember 2009 23:21
> To: statalist@hsphsun2.harvard.edu
> Subject: st: Accessing random coefficients in xtmixed
>
> I am running the following model:
> xtmixed depvar indepvar || groupvar: indepvar, cov(unstr) mle
>
> and I want to access the random effects parameters that Stata
> calculates.  For example, I'd like to be able to refer to sd(indepvar)
> in later work.  Normally this is done with _b[indepvar].  In more
> complex cases, one refers to the equation name first, as described
> here (http://www.stata.com/support/faqs/stat/streg.html).  Normally,
> one can find the equation names by typing "matrix list e(V)" or
> "matrix list e(b)."
>
> However, Stata does not seem to list the standard deviation for
> indepvars in these matrix lists.  That is, the random part of the
> coefficient on indepvar, which is listed prominently in the Stata
> output under Groupvar:  Unstructured (it appears as sd(indepvar)), is
> not shown anywhere in e(b) or e(V).
>
> Is there any way to access/reference the estimated random-effects
> parameters in xtmixed?
> *
> *   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/

*
*   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/
```

 © Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index