Jihad:
please note that, while creating a regression model with no omitted variable bias (no endogeneity) and no heteroskedasticity (quasi-extreme multicollinearity is, in general less relevant for predictors and almost immaterial for controls) means fulfilling regression requirements, searching for statistical significance as your ultimate research goal is not.
Everyone with an average knowledge of frequentist statistics (like me, for instance) does not care that much about p<0.05, but look to confidence intervals, sample size, missing values and the reperesentation of the data generating process provided by the regression model, instead.
That said:
- your first model, as already expressed, due to its sky-rocketing R-sq values looks like a data make-up;
- as far as your second model is concerned: why did you start with -regress- instead of -xtreg-?
please note that, while creating a regression model with no omitted variable bias (no endogeneity) and no heteroskedasticity (quasi-extreme multicollinearity is, in general less relevant for predictors and almost immaterial for controls) means fulfilling regression requirements, searching for statistical significance as your ultimate research goal is not.
Everyone with an average knowledge of frequentist statistics (like me, for instance) does not care that much about p<0.05, but look to confidence intervals, sample size, missing values and the reperesentation of the data generating process provided by the regression model, instead.
That said:
- your first model, as already expressed, due to its sky-rocketing R-sq values looks like a data make-up;
- as far as your second model is concerned: why did you start with -regress- instead of -xtreg-?
Comment