OVTEST interpretation

[Aishwarya Nahata]

Dear Statalist,

I ran my regression

reg ln_trade contig comlang comcol rta intl ln_dist

Thereafter I ran teh OVTEST to check for model specification errors.
the result I got was:

Ramsey Reset test using powers of teh fitted values of ln_trade

H_o : model has no omitted variables

[INDENT=2]F(3, 19571) = 58.61[/INDENT][INDENT=3]Prob > F = 0.000[/INDENT]
While my P-value is less than the minimum threshold of 0.05, I am not sure how to interprest the overall results. Can you please help me interpret this.

Thank you so much.
AIshwaya

[Carlo Lazzaro]

AIshwaya:
short answer: your model is misspecified.
Long(er) reply: with no other details (ie, data example/excerpt poster via -dataex-) it is difficult (for me, at least) to reply more positively.

[Eric de Souza]

The null hypothesis for the RESET test is that the estimated model is correctly specified as to functional form given the variables included in the regression.
The alternative is quite vague: the estimated model is not correctly specified. No specific form is postulated.
Note that the RESET test only tests whether the right functional form has been used for the variables included in the regression.
It does not test for the absence of other variables. In this sense, it is not a general omitted variables test even though Stata refers to it as the -ovtest- .

[Joao Santos Silva]

Adding to the excellent advice already provided, note that (as far as I know) the test is performed with plain vanilla OLS standard errors and therefore is not valid when you need robust or clustered standard errors. This command really needs an update...

Best wishes,

Joao

[Emma Kemp]

Hi all,

I have just stumbled across this forum and was wondering if I could please ask a follow-up question?

I am conducting research on longitudinal panel data and am first running an OLS regression as a benchmark before then moving to panel fixed effects regression. I was hoping to use the ovtest to prove that the OLS model was misspecified and that omitted variable bias likely exists in the form of unobserved individual heterogeneity thereby motivating re-estimation using fixed effects.

Is this an appropriate use of the ovtest or have I misunderstood its purpose?

Thank you in advance for any guidance you can give- I am very new to statistics.

Best wishes,
Emma

[George Ford]

ssc install resetxt

it apparently does Ramsey's Reset test (ovtest) on panel data.

[Carlo Lazzaro]

Emma:
see also -linktest-, whose procedure you can replicate by hand when switching from -regress- to -xtreg,fe-.