what is the best strategy in panel data analysis?

[kelaynak]

Hello,
I will deal with panel data analyis in my thesis. What is the best strategy for it?
My data will be unbalanced panel, most probably non stationary data. I am planning to use Eviews 6.

Firstly,
1) I am planning to check unit roots for all variables (should I use Fisher PP or Fisher ADF for unbalanced panel or IPS?)
2) if there is unit root, I will do panel cointegration tests for only the variables that I have unit roots
3) if there is a panel cointegration, I have two option; taking first difference to remove cointegration (loosing some data and some economical meaning) or ECM Or trying different deterministic trends in panel cointegration test hoping not to have cointegration
4) Panel estimation with GMM (is it right method?)
5) Hausman test for checking whether my model have fixed or random type?
6)normality test for results to check whether residual in normal distribution or not?
7) looking probalities in model eqn to check whether variables statistically significant or not
looking J value to check that the model is sginificant or not in overall
9) looking Durbin Watson for serial correlation( In a book I read that it is not valid in panel data. If not , what is the test for autocorrelation in panel data with GMM?)
10) I think I should look also heterokedastcity but, I could not a test for it in unbalanced panel with GMM.

What else should I do? Is there other tests that I can use in GMM type panel analysis?
Is it right strategy or need any modifications?

I really apprecite for your help. I have looked several books including Baltagi(2005). There is no step by step approach. In all books, there is alot of information for each method, but no strategy for panel analysis from beginning to end.

Thanks in advance
Kelaynak

[akhmad_t]

Dear all,

I face the same problem. I hope someone to solve the problem, please.

Thank you

[seral]

hi all, i have seen your message just now, and i saw that u have wrote it for long time ago,

now i am dealing with the same problem as yours, so probably you r more experienced right now, since i suppose you have finished your thesis, so could you please tell me the steps that you have followed.

Well i hope on the other side that you are still a member of this website.

Thanks for your help in advance