[R] binomial lmer and fixed effects

[Dimitris Rizopoulos]

AFAIK likelihood ratio tests are preferable, especially when you're 
interested in testing several fixed-effects simultaneously. However, 
since in GLMMs the likelihood cannot be calculated explicitly one 
could wonder how does this affect LRTs.

Adaptive Gaussian quadrature is know to provide the best 
approximation; however, taking into account the random-effects 
structure in your model I doubt if it'd ever convergence in this year. 
Thus, I think Laplace is the best you can have in a reasonable 
computing time.

Regarding `method = "ML"' I think this refers to the linear mixed 
model case where you also have the option for REML (the default).

I hope it helps.

Best,
Dimitris

----
Dimitris Rizopoulos
Ph.D. Student
Biostatistical Centre
School of Public Health
Catholic University of Leuven

Address: Kapucijnenvoer 35, Leuven, Belgium
Tel: +32/(0)16/336899
Fax: +32/(0)16/337015
Web: http://med.kuleuven.be/biostat/
     http://www.student.kuleuven.be/~m0390867/dimitris.htm


----- Original Message ----- 
From: "Martin Henry H. Stevens" <hstevens at muohio.edu>
To: "R-Help" <r-help at stat.math.ethz.ch>
Sent: Friday, June 09, 2006 3:55 PM
Subject: [R] binomial lmer and fixed effects


> Hi Folks,
>
> I think I have searched exhaustively, including, of course R-help 
> (D.
> Bates, S. Graves, and others) and but I remain uncertain about
> testing fixed effects with lmer(..., family=binomial).
>
> I gather that mcmcsamp does not work with Do we rely exclusively on 
> z
> values of model parameters, or could we use anova() with likelihood
> ratios, AIC and BIC, with (or without) method="ML" (with didn't seem
> right to me)?
>
> I also received an error using adaptive Gaussian quadrature (but not
> Laplacian approximations):
>
> > mod2 <- lmer(yb ~ reg*nutrient*amd +
> +             (1|rack) + (1|status) +
> +             (1|popu) + (1|popu:amd) +
> +             (1|gen) + (1|gen:nutrient) + (1|gen:amd) +
> +              (1|gen:nutrient:amd),
> +             data=datnm, family=binomial, method="AGQ")
> Error in lmer(yb ~ reg * nutrient * amd + (1 | rack) + (1 | status) 
> +  :
> method = "AGQ" not yet implemented for supernodal representation
>
> I would really appreciate any and all thoughts or leads.
>
> Cheers,
> Hank Stevens
>
> > version
>                _
> platform       powerpc-apple-darwin8.6.0
> arch           powerpc
> os             darwin8.6.0
> system         powerpc, darwin8.6.0
> status
> major          2
> minor          3.1
> year           2006
> month          06
> day            01
> svn rev        38247
> language       R
> version.string Version 2.3.1 (2006-06-01)
> >
>
>
>
> Dr. M. Hank H. Stevens, Assistant Professor
> 338 Pearson Hall
> Botany Department
> Miami University
> Oxford, OH 45056
>
> Office: (513) 529-4206
> Lab: (513) 529-4262
> FAX: (513) 529-4243
> http://www.cas.muohio.edu/~stevenmh/
> http://www.muohio.edu/ecology/
> http://www.muohio.edu/botany/
> "E Pluribus Unum"
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide! 
> http://www.R-project.org/posting-guide.html
> 


Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm