Panel diagnostics/weights

[jbashute]

I have run some preliminary regressions of my sovreign bond spread/credit rating fixed country effects model

spread= c+ spread(t-1) + interest rate + rating dummy

However, I am unsure which weights/covariance matrix to use. How would I go about deciding which, if any, to use, given that there are no heterosk/correlation diagnostic tests available for panel data Eviews. for example, I think there will be heterosk, because the spread chaanges for say greece or portugal are much larger and more irregular than those of austria or the netherlands. there will be outliers for sure. Is one to use the options which produce to largest standard errors? the smallest? or the most significant p values? My model is fairly simple but this is really holding up my research. Thanks in advance.