|
|  |
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: -streg- & equality of coefficients across independent areas
|
From |
"Austin Nichols" <[email protected]> |
|
To |
[email protected] |
|
Subject |
Re: st: -streg- & equality of coefficients across independent areas |
|
Date |
Tue, 6 Jun 2006 00:21:09 -0400 |
Priscila--
I don't use -streg- myself, so I can't speak to any tricks you might
be able to use to run the test you have in mind without running the
model with a dummy and interactions on the union of the two samples.
But I will note that the help file for -streg- specifies
maximize_options; ...
from(init_specs);
(see -help maximize-), so you could specify an initial parameter
vector that allows your regression to iterate only once (or close to
it), given that you have run the regression on both samples
successfully, and can therefore specify the maximum-likelihood
parameter vector. You could -set trace on- to see what the program is
doing if you are concerned that it is taking too long, though that
requires you stop it working and then restart it, of course.
--Austin
On 6/5/06, Priscila Ferreira <[email protected]> wrote:
I understand what you mean now, Austin. But it is literally 'doesn't
run' , hours spent in exactly the same point. No error messages,
nothing. I concluded (perhaps wrongly) that the system could have had
entered in some kind of endless loop.
I have acquired the 64 bits version quite recently, but haven't yet
installed it. In the meantime, I wish I could do the test.
If you have an idea about how to perform the test, no matter how less
straightforward it can be (I'm ready to work on that) , I would be
really grateful.
Priscila
On 05/06/06, Austin Nichols <[email protected]> wrote:
Priscila--
I would not claim that the use of dummy variables is really the only
way to perform the test you want in Stata; but it's the most
straightforward way most of the time. My point was only that a phrase
like "the regression doesn't run" is unlikely to elicit helpful
responses, unless you relate the specific error messages or
constraints you are hitting. I mentioned two ways to relax a
constraint (more memory; Stata MP) but it is not clear to me that
either would help, in the absence of more info about why "the
regression doesn't run." However, I will note that you might need a
64-bit version of Stata to use more memory.
--Austin
On 05/06/06, Priscila Ferreira <[email protected]> wrote:
> Thanks, Austin.
>
> I am currently using Stata 9.2 SE for Linux on a 32 bits processor and
> have 2.5 gb of ram memory available. My model doesn't run in this
> particular machine. It might be the case that if I use another
> architecture the model will run.
>
> So, Austin - do you mean that the use of dummy variables is really the
> only way to perform the test I want in Stata?
>
> Priscila
On 05/06/06, Austin Nichols <[email protected]> wrote:
Priscila Ferreira--
What do you mean when you say
"I suppose
> because the number of variables and observations is too big, the
> regression doesn't run."
precisely?
Do you mean you need more memory or a newer version of Stata that
supports multiple CPUs, or do you mean the regression is infeasible on
any machine?
On 6/4/06, Priscila Ferreira <[email protected]> wrote:
> Good afternoon,
>
> After using -streg- with shared frailty, I am trying to test the
> equality of regression coefficients that are generated from the same
> specification (same dependent variable, same vector of regressors),
> estimated on two different samples. But am not able to.
>
> I tried to implement StataCorp's FAQ:
> http://www.stata.com/support/faqs/stat/testing.html, but I suppose
> because the number of variables and observations is too big, the
> regression doesn't run.
--
Priscila Ferreira
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
