[R-sig-ME] prediction for glmer objects

burg4401 at uni-trier.de burg4401 at uni-trier.de
Mon Mar 2 15:01:59 CET 2009

Hola Virgilio,

thanks for your reply. This indeed helps. Partitioning the data.frame along 
the top-level random effect decreases the magnitude of the model matrices by 
far, such that RAM comes quite handy.

What I'm trying to do is just to replicate the structure of Y from one dataset 
to an other. Though the predictive check was meant in the beginning of his 
work to test whether the models holds on the same data-set. But this is 

Are you coming to the Rhine River Cruise?

Thanks again,

Am Sonntag 01 März 2009 18:02:41 schrieb r-sig-mixed-models-request at r-
> Message: 2
> Date: Sun, 01 Mar 2009 14:29:39 +0100
> From: Virgilio Gomez Rubio <Virgilio.Gomez at uclm.es>
> Subject: Re: [R-sig-ME] prediction for glmer objects
> To: burg4401 at uni-trier.de
> Cc: r-sig-mixed-models at r-project.org
> Message-ID: <1235914179.7168.9.camel at Virgilio-Gomez>
> Content-Type: text/plain; charset="UTF-8"
> Hi Pablo,
> Hope all is fine in Trier. :)
> El dom, 01-03-2009 a las 13:16 +0100, burg4401 at uni-trier.de escribi?:
> > Dear list-users,
> >
> > in my Diploma-Theses I have the need for predicting new Y from a fitted
> > object on a different dataset. Essentially what i want to do is:
> >
> > 	Y ~ X\beta +Zb
> > 	and using the \beta and the variance Components of the random effects to
> > get Ynew ~Xnew\beta + Znew rnorm(b,0,sd(b))
> Is that a sort of predictive check? If you have some many registers, you
> have several options:
> - You can put all your data in a database and then extract it in small
> groups, perform the prediction on that and the add the predictions to
> the database.
> - If you have a big data frame with your data, you could use package
> snow (or snowfall) on a cluster. However, I believe that this uses
> shared memory so you will need a big machine.
> Hope this helps.
> Virgilio

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