Skip to main content
4 events
when toggle format what by license comment
Nov 18 at 20:38 comment added Thomas Lumley glm isn't designed for GLS specifically, it's for generalised linear models, but basically yes. You pass it a vector that is $W$ or proportional to $W$ and it will compute $\hat\sigma^2$. If you want GLS you might try nlme::gls
Nov 18 at 15:43 comment added s5s Let, $cov(\hat{\beta}) = \sigma^2 (X'V^{-1}X)^{-1} = (X'W^{-1}X)^{-1}$. I am passing $W$ - are you saying the library expects me to pass $V$ and it will multiply by the estimated $\hat{\sigma}^2$
Nov 18 at 15:37 comment added s5s I'm a bit confused. How is this different from my equations in which $\sigma^2 = 1$?
Nov 18 at 2:21 history answered Thomas Lumley CC BY-SA 4.0