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: likihood function is not concave in clogit model


From   Nina YIN <yinnina.ccer@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: likihood function is not concave in clogit model
Date   Fri, 29 Jun 2012 14:23:12 +0200

Hi Maarten,

Thank you for your suggestions!
As you suggested, I tried use products and year fixed effect
separately, however, it says that the year dummies omitted because of
no within-group variance.
The year dummies is the same as individual attributes which cannot be
added in this type of model, right?

best,
Nina

On Fri, Jun 29, 2012 at 12:32 PM, Maarten Buis <maartenlbuis@gmail.com> wrote:
> On Fri, Jun 29, 2012 at 10:52 AM, Nina YIN wrote:
>> I have a purchasing data with more than 10000 observations over 20
>> products in 14 years. I am going to estimate the demand in conditional
>> fixed logit model. To solve the endogeneity problem, I need to add
>> product-year fixed effects. which implies 170 fixed effects should be
>> included in the clogit estimation. The results turn out that estimates
>> couldn't be achieved like this:
>
>> If I simply use products fixed effects, the convergence could be
>> achieved and estimates will return. I just don't get what's the
>> problem in my model, several references could achieve convergence even
>> with much higher dimension of fixed effects. But why I couldn't? Is it
>> because of the data variation or my model specification?
>
> By estimating a fixed effects model you throw away a lot of
> information, so it can easily happen that even with 10000 observations
> there just isn't enough information present in your data to estimate
> the effects of a 169 indicator variables. What about adding indicator
> variables for product and year but not their interaction? That way you
> would end up with a more manageable set of 32 indicator variables.
>
> -- Maarten
>
> --------------------------
> Maarten L. Buis
> Institut fuer Soziologie
> Universitaet Tuebingen
> Wilhelmstrasse 36
> 72074 Tuebingen
> Germany
>
>
> http://www.maartenbuis.nl
> --------------------------
> *
> *   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/



-- 
Best Regards,
Nina YIN


Toulouse School of Economics
Manufacture des Tabacs
21 Allee de Brienne
Toulouse, 31000, France
Tel: 0033-(0)5 6123 8348
*
*   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 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index