Hi Gerda,
Since glmmPQL works by repeated calls to lme, it should work the same as lme does, i.e. the default random-effects parameterization is log-Cholesky with an unstructured covariance matrix. Specifying other covariance structures should work the same as in lme, e.g. random=list(grouping=pdDiag(~covariates)) for diagonal; see ?pdClasses.
Good luck,
Cesko
-----Oorspronkelijk bericht-----
Van: R-sig-mixed-models [mailto:r-sig-mixed-models-bounces at r-project.org] Namens Gerda B?rner
Verzonden: vrijdag 2 februari 2018 9:58
Aan: r-sig-mixed-models at r-project.org
Onderwerp: [R-sig-ME] What is the default covariance structure in the glmmPQL function (MASS package)?
Hello,
I sent my query to the r-project.org mailing list first where I got told to rather send it to the r-sig-mixed-models mailing list. I hope someone can help me with my question although the r-sig-mixed-models mailing list refers to issues with using lme4 rather than the glmmPQL function from the MASS package.
So here is my question:
Currently I am trying to fit a generalized linear mixed model with binary outcome using the glmmPQL function in the MASS package.
I was wondering, which variance-covariance structure the glmmPQL function is using by default and if it is possible to vary the variance-covariance structure with the glmmPQL function. Unfortunately I couldn't manage to find out myself.
If it is possible to change it, could someone tell me how to do so? I am especially interested in a diagonal structure versus an unstructured variance-covariance structure.
Thanks a lot in advance.
Gerda
_______________________________________________
R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models