Run Pooled-Fixed-Random Effect Regression

[Khaled Ali]

Dear All
If i have a data set for one dependent variable, 3 independent variables and 4 control variables and my model to be tested as follow:

Dti – D (t-1)i = α0 + cα1 Eti +cα2 Eti*Dinst +cα3 Eti *Dman – cD(t-1)i+ α4SIZE+ α5 LEV+ α6 MTBV+ α 7FCF + uti

Question : can i run regressions if the dependent variable and some of the independent variables are not normal? and if yes, it is will affects the estimation results for the model?

Thank you all in advance

[Carlo Lazzaro]

Khaled:
1) yes, you can run an OLS if the dependent variable and some of the independent variables are not normal, as normality is a residuals distribution requirement;
2) if the regression model is well specified and heteroskedasticity (if present) is addressed via robustified standard errors, your coefficients are ok.

[Phil Bromiley]

Carlo's answer is absolutely correct for the question you posed. However, from your notation and topic description it looks like you may have panel data. If so, you may want to look at xtreg.

[Khaled Ali]

Thank you all for the information and assistance.

Best Regards