Please let me recommend our e-poster:

Poster A Visual Framework for Longitudinal and Panel Studies (with ...

I think the first step it to visually confirm that you have no data issues.

can refer to http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.466.8012&rep=rep1&type=pdf

Dear Umair,

"The Oxford Handbook of Panel Data," edited by Badi Baltagi, Oxford University Press, 2015 shows a number of tests you can do. Some of them listed in the index include constant-correlated effects, cross-section dependence, incidental parameters, moment conditions, noncointegration, overidentifing restrictions, residual based, sphericity, state dependence, unit roots etc.

Data facts are important in any analysis. So diagnostic tests are used to establish data facts since we may not know the data generating process and so we use diagnostic testing to infer to the data generating process

Generally, when panel data is used, test such as unit root test is necessary. Try to download a sample paper and follow its steps by using a tool as Eviews.

Dear Umair,

You say you have data for 6 years, but nothing about the number of panel units (countries?, regions?, sectors?) - how many are they? You need to clarify N=? (with T=6) and then to check which diagnostic tests you may use. Just have in mind, most of the tests about the residuals are asymptotic, so with a very short time series (T=6) they are not reliable. If N is large, you can see what diagnostic tests you can apply in the framework "large sample / short time series".

And, generally, find a suitable book to study the subject matter - I would recommend Wooldridge (beginner: Introductory Econometrics; advanced: Econometric Analysis of Cross Section and Panel Data). Wishing you success!

Dear Venelin,

I have 14 companies and financial ratios for 6 years and i intend to use GMM. for diagnostic tests of autocorrelation and heterosk i believe they are not necessary since GMM is NOT an OLS? is that correct? i have checked multicollinearity and lets say i do check for heterosk and autocor and they are present, they will only make the case of application of GMM stronger since GMM works better in presence of heterosk and autocorr, now coming back to my question since these diagnostic tests will not add value are there other tests that are applicable? if so then how will they add value? please forgive me if i have made any conceptual blunder i will be honored for the correction

So, you have also small sample - I cannot tell you how to proceed in this situation. By my opinion, such powerful methodologies like GMM and many diagnostic tests require T to be large (some tests will be happy if N is large, but it is not too). I would recommend that you check in the basic literature, what you can test at all in this situation.

Alright, But is this correct that Heterosk and Autocor - These diagnostic tests do not apply to GMM, and Data for GMM has its own DIagnostic Tests?

Unfortunately, my suggestion is - with this data - to forget about GMM or other "magical" methods. Just focus on the classics - you can inform yourself about various tests from here:

http://dss.princeton.edu/training/Panel101.pdf

I Did, But had to rule out simple OLS due to apparent endogeneity, Fix/Rand E offered help, but still did not find significant relationship, in addition, Rho Values and Correlation indicated presence of endogeneity, Then and only Then i had to resort to GMM/2SLS,3SLS- Since my data is secondary, 2SLS,3SLS is not applicable.

Now, the researcher's panel is constantly asking about diagnostic testsm, Some have said, You dont need to check Diagnostic tests (Heterosk and Autocor), GMM has its own diagnostic tests Hansen and Sargan. This is where i am confused,