Panel, PCSE, and tests for heteroskedast. & autocorrelation

[alexinusa]

Love Eviews, and its manual is for the most part self-explaining. Great! Together with the highly practical "Econometrics for Finance" (by C.Brooks; a lot of verbal step-by-step explanations, with images of Eviews windows) it enabled me to rapidly advance on my PhD thesis in economics, doing cross-section and time-series regressions, although I unfortunately had not much econometrics at college.

However, when it comes to panel analyses and tests beyond the inital questions of fixed vs. random effects,
Eviews and its manual are much less intuitive/informative (which I understand is partly of coursedue to the more complex material).

So my questions left are:
Before applying the Becks and Katz (1995) PCSE method to my OLS-panel (21 nations, 26 years), I should first test for autocorrelation and heteroskedasticity.
In a panel, how do I
1) test for cross-sectional correlation (e.g. using a Breusch-Pagan LM test),
2) test for autocorrelation (e.g. using a LM test for AR1) (or can I just save the residuals, and test them for a unit root?)
3) test for groupewise heteroskedasticity (e.g. using a Wald-Test)?

Also, practical help would be great for the four PCSE options ("coef covariance method"):
4) cross-section vs. period
each combined with either
5) SUR vs. weight.

Are there any intuitive answers/rule of thumbs? Are these four methods linked e.g. to fixed time and/or fixed country effects? Are they still OLS, or GLS? Does the "Eviews illustrated" book by Startz answer theses questions?

Help on these issues would be really great! :)
Thank you,

Alex

(I understand I might be asking too much, but reading Baltagi, Greene and others (except the book mentioned above) is too complicated for me, especially when it comes applying their maths/matrices practically with Eviews.

I also looked in the forum for answers; it seemed to me others are also looking for these answers to their panel analysis
questions.

So, a big big wish of mine to the QMS team would be to include the above tests directly in Eviews so that we could perform them non-manually. I was hoping Eviews 7 would do the job, as panel regressions are so strongly on the rise, at least in economics.

Or, if Eviews 7 won't include these auto-options yet - maybe you could deepen the manual in this concern? )