Re: st: prob>chi2 in probit results

[Maarten Buis <maartenlbuis@gmail.com>]

On Tue, Jan 29, 2013 at 8:24 PM, nola l wrote:
> Is it correct that as my results showed Prob > chi2     =     0.6615
> which is greater than 0.05, then my model is not a good model and I
> could not really use it?

The null hypothesis of that test is not "my model is not a good model"
but "all coefficients except the constant are equal to 0", and in this
case you cannot reject that hypothesis. This does not mean that all
coefficients except the constant are 0, but only that there is
insufficient evidence in your data to reject that hypothesis; an
absence of evidence is not the same as evidence of absence. However
this is a result that does worry me, and would lead me to consider the
variables involved very carefully, and I would probably end up using
less variables in my model.

> Is it correct that in order to have a acceptable model, I have to get
> prob>chi2 <0.1?

No, as I said above: the null hypothesis of that test is not "my model
is unacceptable". Even if it were, not rejecting that test would not
proof that the model is unacceptable.

> Is there any other command that I can use to test my probit model is
> usable or not?

I would not look for statistical tests, they are typically not that
helpful when selecting a model. There is discussion of various
diagnosics in

http://www.stata.com/bookstore/generalized-linear-models-and-extensions/

and

http://www.stata.com/bookstore/regression-models-categorical-dependent-variables/


Hope this helps,
Maarten

---------------------------------
Maarten L. Buis
WZB
Reichpietschufer 50
10785 Berlin
Germany

http://www.maartenbuis.nl
---------------------------------
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/