Martin Weiss replied to Laura Platchkov:
> In this case, Stata does state the null hypothesis. So if you are willing to
> be wrong at least 2 out of 100 times while interpreting this test statistic,
> reject the null (which does not necessarily mean acceptance of the
> alternative). If you are only willing to be wrong 1 out of 100 times or
> less, don`t. Bottom line: There is nothing specific about this test that
> would distinguish it from any other test.
In addition to Martin's comments, it's worth noting exactly what
-ovtest- tests for.
> "Raymsey test using the powers of the fitted value of the dependent
> Ho: model has no omiited variable
>
> F (3, 168) =3.43
> Prob > F = 0.0169
Many scholars incorrectly interpret a non-significant F ratio to
support the conclusion that we have included all the relevant
variables, when it actually means that there are almost certainly no
omitted _nonlinear_ variables to include on our model.
In your case, F is significant, so you may well have omitted some
quadratic, cubic or otherwise nonlinear variables (or, indeed,
nonlinear transformations of your existing variables) in your model.
Once you've included them and fit the model, running -ovtest- again
may produce a different outcome.
--
Clive Nicholas
[Please DO NOT mail me personally here, but at
<[email protected]>. Please respond to contributions I make in
a list thread here. Thanks!]
"My colleagues in the social sciences talk a great deal about
methodology. I prefer to call it style." -- Freeman J. Dyson.
*
* 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/
