[R-sig-ME] another gls question

[Mark Leeds]

Thanks Karl for the reference. I'll check it out.

Mark

Thierry:  In my last email, I meant L prime L = Omega inverse not Omega.
Essentially,
you find a transformation such that the error term becomes independent and
you
can use the Gauss Markov Theorem because the model has been transformed to
an OLS model. Any decent econometrics book will explain this approach way
more clearly than I currently am :).

My best guess is that, since the help file for gls refers to "Ruppert and
Carroll" as the
main reference, ( I've ordered it. If it's useful, I'll let this list know
), the gls function probably doesn't use the" Transform to  OLS" approach.
Therefore, the "normalized" residuals are referring to something else
rather than the OLS residuals.

Thanks again to both of you for your help.












On Fri, May 8, 2015 at 6:19 AM, Hufthammer, Karl Ove <
karl.ove.hufthammer at helse-bergen.no> wrote:

> Thierry Onkelinx:
> > SSTOT = SSREG + SSTOT is relation that holds with a lineair model with
> OLS.
> > I'm not sure if it still holds with gls.
>
> For a R² measure for GLS models, see this paper:
>
> A. Buse (1973): Goodness of Fit in Generalized Least Squares Estimation
> The American Statistician, Vol. 27, No. 3, pp. 106-108
> http://www.jstor.org/stable/2683631
>
> --
> Karl Ove Hufthammer
>

	[[alternative HTML version deleted]]