Dear All,
Hello,
First of all, i hope that you all and your families are doing well because of the pandemic. Hopping that the days to come will be much better.
I'm a PhD student in management sciences, and for my study i'm using an econometric model (panel data model). Actually, i'm studying the case of two countries, for 18 years and i have 31 observations.
I followed a paper studying same topic and used same model, but with less countries. So i used FE model as the basic one and pooled OLS & GMM for robustness.
But i want to ask a question about stationarity, i know that when we have (T>N), we use the non stationary panel data analysis such as Fully Modified OLS (FMOLS), Dynamic OLS (DOLS), Pooled Mean Group (PMG) etc. And when we have (N>T), we apply the traditional panel data analysis such as fixed effect(FE), random effect(RE), pooled OLS(POLS) etc.
But in my case even if i have T>N i used the traditional panel data, as basic one i used FE model because it control ou remove the unobserved time invariant variables. But before all this i used the unit root test.
The question that i want to ask is related to stationarity, when we use FE, OLS or GMM should we do the unit root test or its irrelevant ?
Actually, my variables are most of them stationary at level I(0) (because three of my variables are not stationary at level, but they all are stationary at first difference, so because i have them all stationary at first difference, i used Pooled OLS, i wanna know also, if what i did is correct ? When we have all the variables stationary at first difference, can we run Pooled OLS ?
If yes, do you have any reference saying that when we have all variables stationary at first difference, we can use pooled OLS.
If you can share your response or advice i would be very grateful.
Thank you in advance.
Best regards
Fatima
Dear Fatema Matefa,
I hope you are doing well.
Here, I will outline all the steps that need to be taken during panel data analysis.
1. Check the Cross-sectional Dependence.
There are certain tests for cross-sectional dependence, the Lagrange multiplier (LM) test formulated by Breusch and Pagan (1980) is suitable for your study, because N
Dear Dr. Aqib,
Im fine alhmd, thank you. I hope same for you inchAllah.
Thank you for your response. Appreciated.
But have a question related to the step one, you said that : "the Lagrange multiplier (LM) test formulated by Breusch and Pagan (1980) is suitable for your study, because N
Dear Fatema,
I have attached some papers. Please read carefully, I hope that these papers will answer all your questions.
Yes, Stationarity is necessary many useful analytical tools and statistical tests and models in your research rely on it.
Only carry out pooled OLS if the data is pool-able otherwise proceed with fixed effect regression. So decide between random effects and fixed effects model coefficients.