[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: testparm

From   Maarten buis <>
Subject   Re: st: testparm
Date   Tue, 4 Mar 2008 13:20:43 +0000 (GMT)

--- Chiara Mussida <> wrote:
> I know that I can't obtain causal interpretations,and i'm not
> interested in such a kind of interpratations. I simply want to see
> whether my covariates are significant and if there are relevant
> differences in their behaviour. As regards my interest, I want simply
> to be sure that the appropriate choice for me is to run separate
> models. From the pooled I lbtain coef stat different for the
> geographical partitions and it is well known that the unemployment
> assumes remarkable different regional only
> desire is to be sure (econometrically) about this!!
> chiara

Unfortunately, model selection is more an art than a science. There are
several things you can do. Some people like tests, but there is a big
problem that it changes the logic behind the subsequent test especially
if the test has low power, like a test of many interaction effects. So
in your case I would not do that. Others like graphs of for instance
residuals, or comparing BIC or AIC across models. However, you will
never get even close to sure, econometrically or otherwise.


Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands

visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434

+31 20 5986715

Rise to the challenge for Sport Relief with Yahoo! For Good
*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index