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: non-linear models not converging


From   Nick Cox <njcoxstata@gmail.com>
To   "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
Subject   Re: st: non-linear models not converging
Date   Wed, 22 May 2013 14:11:32 +0100

-nl- does not use maximum likelihood. -xtpoisson- does.

As you did not specify -xtpoisson- in your first post, none of us
could write comments geared to that command.

You can reach through -xtpoisson- to specify initial values. See the
help and click on maximize_options.

But you have extra options, including changing -tech()-. For example,
I find -tech(bhhh)- works wonders, and I have no precisely zero idea
why. But I like it nevertheless.

However, no trick is guaranteed to work. At a rough guess from
Statalist postings, most models that don't converge in Stata are just
a bad idea and there is no easy fix for that.
Nick
njcoxstata@gmail.com


On 22 May 2013 13:59, James Bernard <jamesstatalist@gmail.com> wrote:
> Thanks Nick,
>
> I am not writing the MLE. I am using -xtpoisson
>
> How can I supply the initial value for the numeric solution?
>
> Thanks,
> James
>
>
> On Wed, May 22, 2013 at 8:55 PM, Nick Cox <njcoxstata@gmail.com> wrote:
>> I presume focus on -nl-.
>>
>> Convergence is more likely if
>>
>> 1. the model is actually right for the data in a qualitative sense
>> (easy to say, hard to define, obvious when it fits well)
>>
>> 2. you supply good initial guesses for the parameters (this is perhaps
>> the easiest one to tweak)
>>
>> 3. you are estimating a small number of parameters
>>
>> 4. you have a good ratio of data points to parameters
>>
>> 5. the data are not grotesquely behaved (e.g. outliers and high
>> skewness can be just as problematic as with linear models)
>>
>> 6. the model is not highly nonlinear (the textbooks are full of this)
>>
>> 7. I like lists to have about 7 items, so something else belongs here.
>>
>> Maarten Buis should have a Euro for every time he's recommended
>> retreating to a simpler model when a complicated one doesn't converge,
>> and then adding complexity one step at a time. But it's good advice.
>>
>> Nick
>> njcoxstata@gmail.com
>>
>>
>> On 22 May 2013 13:43, James Bernard <jamesstatalist@gmail.com> wrote:
>>> Hi all,
>>>
>>> I understand that the numeric methods used for estimation of models in
>>> Stata (and any other package) may result in a model that does not
>>> converge.
>>>
>>> Do you happen to know of any trick to help make the model converge? To
>>> increase the chance of converging?
>>>
>>> Thanks,
>>> James
>>> *
>>> *   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/
> *
> *   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/


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