Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


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

Re: st: RE: ml and parmby versus statsby


From   Nick Cox <njcoxstata@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: RE: ml and parmby versus statsby
Date   Wed, 12 Dec 2012 18:05:39 +0000

We can be pleased that you no longer see a problem, but what happened
remains mysterious.

1. -parmby- permits other options but these are in turn (if I follow
Roger's code correctly) passed to -parmest-, which does not in fact
allow a -quietly- option. So why your earlier code ran at all is a
puzzle.

2. In general, the effect of a -quietly- option in Stata would be that
something is done "quietly", although that's a convention, not a rule.
If so, results should be exactly as they were before, not different.

I spell out here, as we are all asked to do, that -parmby- is a
user-written command. I looked at -parmby-  from SSC and -which- tells
me

*! Author: Roger Newson
*! Date: 02 September 2009

Nick

On Wed, Dec 12, 2012 at 5:41 PM, S. Suetens <S.Suetens@uvt.nl> wrote:

> This is very helpful.
> I re-ran the parmby command, but without including the quietly option, and estimation results are the same now as those under statsby.

Roger B. Newson

> I don't know how your -statsby- and -parmby- results are different, as you haven't told us. However, -statsby- creates a  dataset with 1 observation per by-group, and different variables for the estimates and standard errors for different parameters. -parmby-, on the other hand, creates a dataset with 1 observation per parameter per by-group, with 1 variable for the estimate and 1 variable for the standard error, and other variables for the confidence limits and P-values.
>
> I am a bit mystified about the -quietly- option for -parmby-. As the author of -parmby-, I never knew that -parmby- had a -quietly- option, although it can take the -quietly- prefix like other Stata commands.

On 12/12/2012 15:52, S. Suetens wrote:

>> Is there anyone who can explain why maximum likelihood estimation gives different results under parmby as compared to statsby?
>> I have run the two commands below (based on the same likelihood function) and they give very different parameter estimates and standard errors.
>>
>> statsby _b _se, by(id) saving("D:\Statsby.dta", replace) nodots: ml
>> model lf MaxLik (lambda: strategy =) (rhosigma: RHO SIGMA, nocons),
>> max difficult
>>
>> parmby "ml model lf MaxLik (lambda: strategy =) (rhosigma: RHO SIGMA,
>> nocons), max difficult", by(id) saving("D:\Parmby.dta", replace)
>> quietly

*
*   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/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index