[R] Pseudo R squared in gls model

[Mike Miller]

On Thu, 23 Aug 2012, Gary Dong wrote:

> I'm wondering if the gls function reports pseudo R. I do not see it by 
> summary(). If the package does not report, can I calculate it in this 
> way?
>
> Adjusted pseudo R squared = 1 - [(Loglik(beta) - k ) / Loglik(null)] where
> k is the number of IVs.


We've been using this R² instead, but I'm not sure if it is available in 
any package:

Buse, A. (1973), "Goodness of Fit in Generalized Least Squares 
Estimation," American Statistician, 27, 106-108.

http://www.jstor.org/stable/10.2307/2683631

Equation 15.  Make sure to compute y-bar correctly (it is the regression 
coefficient in the intercept-only model).

To do GLS, we have been using OLS after transformation (multiplying by the 
"T" matrix in Buse's notation) where T'T = inv(V).  So the equation on the 
final page works for us.

Let me know if you can't access that PDF and I'll send a copy.

Mike

-- 
Michael B. Miller, Ph.D.
Minnesota Center for Twin and Family Research
Department of Psychology
University of Minnesota