Bookmark and Share

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: Warning convergence not achieved


From   William Buchanan <[email protected]>
To   "[email protected]" <[email protected]>
Subject   Re: st: Warning convergence not achieved
Date   Sun, 5 Aug 2012 11:57:40 -0700

Hi Ebru,

How many observations are there in each of the industry indicators?  If there are any industries where there are a relatively small number of observations you may want to consider collapsing that dummy with another group.

HTH,
Billy

Sent from my iPhone

On Aug 5, 2012, at 10:42, Ebru Ozturk <[email protected]> wrote:

> Hi Billy,
> 
> Thank you. But what do you mean with the cell sizes? Actually I have a few dummy variables in the model and also I include 9 industry dummies. When I add industry dummies, the model does not work.
> 
> Ebru
> 
> ----------------------------------------
>> Subject: Re: st: Warning convergence not achieved
>> From: [email protected]
>> Date: Sun, 5 Aug 2012 09:02:49 -0700
>> To: [email protected]
>> 
>> Hi Ebru,
>> 
>> Have you checked the cell sizes for your dummy variables? If you have one or two dummy variables that make up a small percentage of your sample size you may want to collapse them into a single category. This type of issue happens often and Maarten Buis has given this advice to numerous folks on the Statalist who are having difficulties with model convergence.
>> 
>> HTH,
>> Billy
>> 
>> 
>> 
>> On Aug 5, 2012, at 8:12 AM, Ebru Ozturk wrote:
>> 
>>> Dear Joao,
>>> 
>>> Thank you. Actually, when I run the Tobit estimation it works very well. But when I add industry dummies (9 industries) into the model it gives the warning. How can I deal with that?
>>> 
>>> Kind regards
>>> Ebru
>>> 
>>> 
>>> ----------------------------------------
>>>> From: [email protected]
>>>> To: [email protected]
>>>> Date: Sun, 5 Aug 2012 15:42:50 +0100
>>>> Subject: RE: st: Warning convergence not achieved
>>>> 
>>>> Dear Ebru,
>>>> 
>>>> The reason may simply be that the Tobit ML estimates do not exist for your
>>>> data set. This will happen, for example, if one of your regressors is a dummy
>>>> that is equal to one only when y is equal to zero. The problem, in a different but
>>>> related context is discussed here:
>>>> http://www.sciencedirect.com/science/article/pii/S0165176510000832
>>>> http://www.stata-journal.com/article.html?article=st0225
>>>> 
>>>> A simple way to check for this is to run a probit instead of a Tobit, with the same
>>>> y and the same regressors, and see if you have a "perfect predictor".
>>>> 
>>>> Hope this helps,
>>>> 
>>>> Joao
>>>> 
>>>> 
>>>>> Dear All,
>>>>> 
>>>>> When I run the tobit estimation, sometimes it gives this warning "convergence
>>>>> not achieved". What could be the reasons? and how can I deal with that?
>>>>> 
>>>>> Best wishes
>>>>> Ebru
>>>>> *
>>>>> * For searches and help try:
>>>>> * http://www.stata.com/help.cgi?search
>>>>> * http://www.stata.com/support/statalist/faq
>>>>> *
>>>> 
>>>> 
>>>> *
>>>> * For searches and help try:
>>>> * http://www.stata.com/help.cgi?search
>>>> * http://www.stata.com/support/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/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/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/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/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index