[Carlo Lazzaro]

Jihad:
you're correct.
Your results are statistically "valid", but should be interpreted in the light of the theorethical background you mention (corporate social responsibility).

[Jihad El Yaagoubi]

Carlo, it doesn't matter if the p-value of test(F) is superior than 0.05 in my model ?

[Carlo Lazzaro]

Jihad:
I consider valid your results even though the -p-value of the F omnibus test tells that there's no evidfence that your model is actually different (or better, can be considered more informative) than the mean of the dependent variable.
That result (that, with a bit of bayesian flavour, I would see an important starting point for future research) can be the offspring of a (too) small sample size.
I cannot say if you can collect more data and/or the literature in your research field reports regression model which resulsts are similar to yours.

[Jihad El Yaagoubi]

And we can say it's significant at 0.1 level? (I'm looking for arguments to defend the model in my report thesis)

[Carlo Lazzaro]

Jihad:
yes, that may be an option to be discussed with your supervisor.

[Jihad El Yaagoubi]

Ok. Thank you very much Carlo for your help ..

[Jihad El Yaagoubi]

Sorry Carlo I ask you again, when reporting results should I report fe , re and pooled ols models and say why the pooled ols is more appropriate. Or directly report the pooled ols results ?

[Carlo Lazzaro]

Jihad:
I would report pooled OLS only, describing the scientific pathway that supported that choice (eg, BP test results).

[Jihad El Yaagoubi]

Thank you Carlo ..

[Jihad El Yaagoubi]

Hi,
in my thesis I report descrptive statistics and correlation table. But i didn t find a significant correlation between CSR (the independant variable) and the other dependant variables. And this is contradictory with a result from my pooled ols model. How can I deal with this? I can't run a regression when there is no correlation ?

[Jihad El Yaagoubi]

The coef of correlation=0.0682.

[Carlo Lazzaro]

Jihad:
in OLS regression, the contribution of predictors to the variation of the conditional mean of the dependent variable is adjusted for the effect of the remaining one.
With correlation, you test how (and if) two (or more) variables varies together (in the same direction; in the opposite durection; in no direction at all).
In sum, I do not think that your results sound contradictory.

[Jihad El Yaagoubi]

Even if the p-value of all correlation coefficents are unsignificant?
and the pooled ols model suggests that there is a relationship.

[Carlo Lazzaro]

Yes, for the reasons given in my previous reply.

[Jihad El Yaagoubi]

Thank you Carlo! Your reply is helpful