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 at the end of May, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: different p-values in mfx after xtprobit


From   Maarten Buis <maartenlbuis@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: different p-values in mfx after xtprobit
Date   Fri, 13 May 2011 09:19:25 +0200

On Thu, May 12, 2011 at 6:51 PM, ramesh wrote:
> Hi, I am using xtprobit model for my research. I want to find the marginal
> effect after the xtprobit model. I used the "mfx compute, predict(pu0)"
> command and got the results but my p-value is totally different and it turns
> out 0.99 for almost all variable. I have attached the xt probit result and
> its corresponding mfx herewith. Is there any suggest to get consistent
> p-value in calculating the marginal effect?

The fact that the p-values are different is to be expected as you are
testing different null hypotheses.

However, the results you show do indicate a problem in your model. One
problem I can see is that you are either modeling an extremely rare
event or you did not center your explanatory variables. I see that
because the constant is -98, which is corresponds to a very very small
probability for the group that has the value 0 on all covariates. If
you are modeling such an extremely rare event than I would say that
the standard error returned by -mfx- accurately represent the amount
of information available in your data. However, I suspect you just
forgot to center your variables. This can be a big deal for this type
of models because what they do is "estimate" group specific constants,
and if you do not center your variables the constants will represent
groups that are way outside the range of your data. This can often
lead to anomalies like the one you are reporting.

So, I would look at each explanatory variable and see if the value 0
represents a meaningful value that is within (or very close to)  the
range of the data. When that is not the case I would pick some
meaningful number for that variable (for example the mean) and
subtract that number from that variable, such that value 0 no
represents that meaningful number.

Hope this helps,
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/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index