EViews.com

Hey there! I appreciate all that you guys do. I have a question with respect to correcting for heteroskedasticity in panel data. I am not sure as to which coefficient covariance method to use i.e. white cross section, white period or white diagonal? I have an unbalanced panel data set with 33 countries over 4 periods of time. I have already corrected for serial correlation by using a manual method that was suggested by Wooldridge. However, I need to correct for heteroskedasticity across the countries. When I use the white cross section option, I receive a warning that states that the estimated coefficient covariance matrix is of reduced rank. When I use the white diagonal option, I do not receive this warning. Which one am I supposed to use, and can you briefly explain what each of the above three option does or at least point me to a reference that explains it?

Thank you!