Test for the need of squared variable specification in fixed effects model?

[Sabrina Muller]

Hello everybody,

I'm running a fixed effects model on panel data and would like to test if I need the squared variable additionally to the normal variable in my model. I would expect the relationship to be u-shaped but the coefficients are significant in both cases. Is there a test to see which model specification is the better one in a fixed effects panel data model? I can't see from the graphs I created.

Thank you for your help

[Clyde Schechter]

The statistical significance of the coefficients is not really helpful in this setting. If you show the Stata output from both models, someone may be able to guide you. Please read the Forum FAQ, with special attention to #12, so you understand how to use code delimiters to properly display the exact outputs.

[Carlo Lazzaro]

Sabrina:
as an aside to Clyde's helpful advice, you might be interested in the following thread: https://www.statalist.org/forums/for...nel-data-model.
Actually, you can't rely on a test that guides you to the right specification of you panel regression model (and that holds for -estat ovtest- after -regression-: the Ramsey test won't tell yoy which variable you forgot to include among predictors, linear, squared or cubic it may be), but only on tests that tell you about the evidence of misspecification of your regerssion model.
Hence, no black or white magic there...