Re: st: Dyadic fixed effects and areg, absorb

[Nick Cox <njcoxstata@gmail.com>]

Sorry, but I don't know -- or shouldn't say even if I did. I don't
even know if this is trade, migration, conflicts or anything else, and
none of those are my field. That's surely a matter for you to seek
advice from your supervisor/advisor if you have one, or the
literature.

Supposedly, John von Neumann said "With four parameters I can fit an
elephant, and with five I can make him wiggle his trunk." I know
software makes it much easier to fit complicated models than whenever
he (may have) said that, but more complicated models are not
necessarily better.

On Fri, Apr 15, 2011 at 4:06 PM, emanuele mazzini
<madsoenistata@gmail.com> wrote:
> Well, I have to say I am a student willing to get a degree (i.e I am
> doing such a research for my final thesis). I asked that because I
> know that some authors use that and insert in their regressions both
> dyadic and year fixed effects. Obviously it is a kind of robustness
> check, so I do not necessarily expect my main variables to be
> statistically significant but if they do, I can say that result are
> definitely robust, aren't they?
> Since you suggested a more parsimonious model, what would it be? What
> kind of fixed effects would you insert in such a regression?
>
> Thank you very much,
> Regards,
> Emanuele.
>
>
> 2011/4/15 Nick Cox <njcoxstata@gmail.com>:
>> It starts sounding like a statistical equivalent of a rather empty
>> model. If I say, interactions between countries over time depend on
>> precisely which countries and precisely when, the reaction to those
>> words is, Well, yes. If you say, and here are thousands of parameters
>> to summarize that, how is that scientifically or practically
>> interesting or useful?
>>
>> I don't know what your overall objective is. Are you a student and you
>> want to pass an examination or to get  a degree? Are you a researcher
>> and you want a publishable paper? Either way, as you ask for
>> reactions, mine is that I would expect anyone in your friend to be
>> more interested in a much more parsimonious model.
>>
>> On Fri, Apr 15, 2011 at 3:31 PM, emanuele mazzini
>> <madsoenistata@gmail.com> wrote:
>>> Dear all Stata users,
>>> I am trying to work out a regression with an unbalanced bilateral
>>> panel dataset in which I need to include both dyadic fixed effects and
>>> year fixed effects. Since my dataset is very huge (it includes more
>>> than 600,000 observations) and I cannot generate a dummy for the
>>> variable dyads (which is the unique dyads identifier), I thought that
>>> I can accomplish this by using the following command:
>>>
>>> areg...., absorb(dyads)
>>>
>>> by including only the dummies that refer to the years to take into
>>> account the year fixed effects, while absorb(dyads) is supposed to
>>> take into account the dyadic fixed effects.
>>>
>>> I am not looking really for an help, but more for an advice: what do
>>> you think about this solution to my problem (i.e. to the impossibility
>>> to generate more than 15,000 dummies)? Do you think this may work?
>>>
>>> Thanks to all of you in advance for your time and consideration,

*
*   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/