[R] A question about corAR1 and grouping

Saptarshi Guha saptarshi.guha at gmail.com
Wed Jan 6 19:55:08 CET 2016


I was under the impression that in a call to gls, if i specify

gls(.., cor=AR1(form= ~ 1 | subject))

and suppose there are N subjects, then i would have obtained estimates
for N Phi (s) (the AR1 coefficient) - but I only get 1 estimate.

I then checked "Linear Mixed-Effects Models Using R" and it i must
have confused it for the weights(e.g. varIdent) where there are indeed
group(strata) specific estimates.

Is there a way to include correlations  and have group specific estimates?


More information about the R-help mailing list