Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: mepoisson vs meqrpoisson
From
[email protected] (Jeff Pitblado, StataCorp LP)
To
[email protected]
Subject
Re: st: mepoisson vs meqrpoisson
Date
Fri, 16 Aug 2013 11:52:58 -0500
Phil Clayton <[email protected]> is using both -mepoisson- and
-meqrpoisson- to fit a 3-level mixed effects model and is getting inconsistent
results from -mepoisson-:
> I have been experimenting with the new -mepoisson- and -meqrpoisson-
> commands and have found substantively different results that I can't
> explain.
>
> ...
> mepoisson inf apd, exp(duration) || hospid: || patient:
> ...
> meqrpoisson inf apd, exp(duration) || hospid: || patient:
> ...
Phil sent us a copy of the data and commands used to reproduce the problem.
We have determined that -mepoisson- is suffering from a likelihood scaling
problem with hospid=53, which contains 527 patients. Phil can eliminate the
likelihood scaling problem by excluding this group of observations, doing so
will yield consistent results between -mepoisson- and -meqrpoisson-, even
between the two datasets (one being a collapsed version of the other).
In diagnosing the problem, we have found a bug in the code that -mepoisson-
uses to compute the mean-variance adaptive parameters. We have corrected the
error and will release this fix in a future update to Stata.
With this fix we have verified that -mepoisson- produces the expected results
when compared to -meqrpoisson- for both datasets.
--Jeff
[email protected]
*
* 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/