st: -ovtest and -linktest in small samples

[Lars12398@web.de]

Dear Statalist,

I have a question concerning model misspecification diagnostics
in small samples with -ovtest and -linktest. My sample consists of 50 countries.
I have seven independent variables and calculate several models for different scenarios. Both tests indicate a
model misspecification for all of them.

I could "solve" the problem for some of the models by transforming a variable (log) and also adding a quadratic term
(log^2) but not for others. Moreover, the tests seem to be sensitive to the inclusion of some countries
in the regression that were not identified as outliers or influential cases by regression diagnostics. For example,
the ovtest becomes insignificant for one of the models without one country in the sample.

I read that it is especially important to pass the ovtest, but I could not find anything in the literature about the
small sample properties of both tests. Are they still essential in order to interpret a regression model or
should one rely more on graphical diagnostics?

Thanks for Consideration.

Lars
___________________________________________________________
SMS schreiben mit WEB.DE FreeMail - einfach, schnell und
kostenguenstig. Jetzt gleich testen! http://f.web.de/?mc=021192
*
*   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/