Weighted statistics for random effects regression

[ralf]

Hi,

it would be great if anybody could help me to better understand the EViews 7 output for random effects models: What is the difference between weighted and unweighted R-squared? The manuals didn't really help much... Can anybody help?

Thanks in advance,

Ralf

[EViews Glenn]

If you look at most textbooks on panel data, you will see that the random effects model is actually a GLS estimator, which is in effect analogous to a weighted least squares problem. EViews reports statistics for the regression run in the weighted data space (after GLS) transformation, as well as the less interesting statistics using the GLS coefficient results but the original Y and X data. The latter are have been requested by a number of users who want to see things in the original data metric, but they have no particular statistical interpretation.