Different within-group error correlation matrices per group
Dear Kevin, I must confess I found both sites on google before asking here, but I could not make much sense out of them. It clicked now that I realised lme models different variances per factor through variance functions and not through the correlation structure. It is even covered in the book I searched on. Just in case it is of help for future readers, the following model
lme(fixed = y~1+time, random = ~1+time|ind, correlation = NULL, weights =
varIdent(form = ~ 1 | group)) has both fixed (intercept and slope) and random (again intercept and slope)
On 28 October 2015 at 14:59, Kevin Wright <kw.stat at gmail.com> wrote:
You might find the following links contain some useful information. http://stackoverflow.com/questions/11819720/converting-repeated-measures-mixed-model-formula-from-sas-to-r Factor-specific variances in R https://rpubs.com/bbolker/6298 Harris wateruse http://www.inside-r.org/packages/cran/agridat/docs/harris.wateruse On Tue, Oct 27, 2015 at 8:57 PM, Luis Damiano <luiggi at gmail.com> wrote:
Dear all, I am working with lme and I would like to have different within-group error correlation matrices per group (\Lambda_i in 5.1 from Pinheiro & Bates). Currently, my sentence looks like the following correlation = corCompSymm(form = ~ 1 | ind) which imposes the SC structure into the within-group error correlation matrix, marking the individuals by "ind". According to the following example in SAS given by an instructor, it is possible to estimate different within-group error correlation matrices per sex using the following sentences (see "group=gender" in fourth line): proc mixed data=dental;
class child gender; model distance = gender gender*age / noint solution; repeated / group=gender subject=child; random intercept age / type=un subject=child g gcorr v vcorr; run;
I cannot figure out how to reproduce this in R. I took a look at the
documentation as well as the first five chapters of Pinheiro & Bates
(2000)
without luck.
Cheers!
[[alternative HTML version deleted]]
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
-- Kevin Wright