Thanks. This makes perfect sense.
-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of JVerkuilen (Gmail)
Sent: Wednesday, October 31, 2012 2:13 PM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: CI for ri variance after melogit
On Wed, Oct 31, 2012 at 1:46 PM, Bontempo, Daniel E <deb193@ku.edu> wrote:
>
> Can someone explain why the nlcom CI gets smaller?
The one generated by -nlcom- is point estimate +/- 1.96*SE. This is the delta method estimate, whereas I suspect what the -xtmelogit- one does is work on the estimation metric, which is much better approximated by the normal, and then transformed back to the scale of a variance. The latter is much more accurate, for the same reason that you would generally not want to form a confidence interval for an odds ratio directly but instead work on the log-odds scale and then transform back.
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/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/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/
