[R-sig-ME] Switch to ML during model simplification

Douglas Bates bates at stat.wisc.edu
Mon Mar 21 17:55:15 CET 2011

On Sat, Mar 19, 2011 at 11:56 AM, Iker Vaquero Alba <karraspito at yahoo.es> wrote:
>    Thank you very much!! With the first question, I mean if you have to change the method when fitting the model or when updating  it. I mean, if you have to write:
>    model1<-lmer(y~a*b*c+(1|random1/random2),REML=FALSE)
>    Or you write the previous without "REML=FALSE" but you write that command when simplifying the model (in the first "update"). Actually, I have tried both, and writing "REML=FALSE" when fitting the model does not seem to have any effect, as when typing "summary(model1)" it always says "model fit by REML method" no matter wether you wrote "REML=FALSE" or not, and the first simplified model is analysed by "REML" or "ML" depending on what you write.

Could you provide a reproducible example of REML=FALSE failing?  Also,
please include the output of


so we know what versions of the packages you are using.

>    Hope it makes sense now. Thank you.
> --- El sáb, 19/3/11, Douglas Bates <bates at stat.wisc.edu> escribió:
> De: Douglas Bates <bates at stat.wisc.edu>
> Asunto: Re: [R-sig-ME] Switch to ML during model simplification
> Para: "Iker Vaquero Alba" <karraspito at yahoo.es>
> CC: r-sig-mixed-models at r-project.org
> Fecha: sábado, 19 de marzo, 2011 10:10
> On Tue, Mar 15, 2011 at 3:29 PM, Iker Vaquero Alba <karraspito at yahoo.es> wrote:
> >
> >    Hello, list members:
> >
> >    I am fitting a mixed effects model with a split-plot analysis. I have read that in order to be able to use anova to compare models with a different fixed-effects structure, it's not correct to use the default REML method, so it must be changed to ML.
> >
> >    - When using "lme", that can be done just writing: method="ML", whereas when using "lmer" you achieve that by writing "REML=false".
> Actually, that's REML=FALSE (case is important).
> >    - My questions are:
> >       1. Is it necessary to specify the ML method when fitting the m

More information about the R-sig-mixed-models mailing list