Hi Scott -
As you and Paul and Marten have suggested, it looks like -nlcom- is the
easiest way to do what I want, however thanks for the tip about _diparm
- never heard of it before - but I like it!
I think I'll pass on rewriting -xtlogit- , but thanks for pointing out
where it can be "fixed".
Al
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Scott
Merryman
Sent: Wednesday, June 11, 2008 11:51 AM
To: [email protected]
Subject: Re: st: accessing delta-method-derived standard errors
I believe that should be -nlcom (exp(.5*[#2]_cons))-
Another way would be
-_diparm lnsig2u, level(`level') label("sigma_u") function(exp(.5*@))
derivative(.5*exp(.5*@))-
where the r(se) will contain the standard error.
Or (using at your own risk),
add the lines
-scalar sigma_se = r(se)- at about line #715 after the -_diparm lnsig2u,
level(`level') label("sigma_u")- line
and
-est scalar sigma_u_se = sigma_se- at about line #674 before the
-
global XTL_madapt - line.
Now:
. sysuse auto,clear
(1978 Automobile Data)
. xtlogit fore mpg, i(rep) nolog
Random-effects logistic regression Number of obs =
69
Group variable: rep78 Number of groups =
5
Random effects u_i ~ Gaussian Obs per group: min =
2
avg =
13.8
max =
30
Wald chi2(1) =
4.62
Log likelihood = -31.540725 Prob > chi2 =
0.0316
------------------------------------------------------------------------
------
foreign | Coef. Std. Err. z P>|z| [95% Conf.
Interval]
-------------+----------------------------------------------------------
-------------+------
mpg | .1462312 .0680197 2.15 0.032 .0129151
.2795474
_cons | -4.397192 1.712402 -2.57 0.010 -7.753439
-1.040945
-------------+----------------------------------------------------------
-------------+------
/lnsig2u | .8701574 1.149525 -1.382869
3.123184
-------------+----------------------------------------------------------
-------------+------
sigma_u | 1.545085 .8880564 .500857
4.766404
rho | .4205076 .2801172 .0708492
.8735078
------------------------------------------------------------------------
------
Likelihood-ratio test of rho=0: chibar2(01) = 7.16 Prob >= chibar2 =
0.004
. disp e(sigma_u_se)
.88805638
Scott
On Wed, Jun 11, 2008 at 11:24 AM, E. Paul Wileyto
<[email protected]> wrote:
> Check out the help on -nlcom-
>
> You would get the exponentiated version of the SE by using :
>
> -nlcom (exp([#2]_cons))
>
> That will transform the constant asscoiated with the second ML
equation.
>
> To grab that value for other things, type -return list-
>
> and you should see the matrices and scalars generated by nlcom
>
> Paul
>
>
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
