Page 1 of 1
Feasible Generalised Least Squares
Posted: Tue Jan 13, 2015 10:59 am
by chettyd3
Hi,
I am estimating a panel data model making use of FGLS with period effects in eviews and would just like to understand how the weighting of the different periods is calculated in eviews?
Re: Feasible Generalised Least Squares
Posted: Tue Jan 13, 2015 6:00 pm
by EViews Glenn
Period effects or period weights?
Re: Feasible Generalised Least Squares
Posted: Tue Jan 13, 2015 10:41 pm
by chettyd3
period weights
Re: Feasible Generalised Least Squares
Posted: Wed Jan 14, 2015 11:31 am
by EViews Glenn
The weighting matrix uses the inverse of the period-specific variances of the residuals from the unweighted model. The balanced Kronecker form of the weight matrix depends on the way you represent your panel, but the easy way to think about it is that (effectively), each observation is weighted by the inverse of the standard deviation (non-df corrected) of that period's residual variance, as computed from the unweighted model residuals.