Hi all, I was working with the "MEMSS" & "mle4" library's under R version 2.15.1. apparently some practical functions of do not work under R 2.15.1.
After searching the archives i found a mail thread on this subject, stating that these problems were partialy solved for "R 2.12.0" but only for "lmer()" not for "glmer()". Is someone aware of an update available of these library's ? Or should I install "R 2.12.0" and "lme4a". Kind regards, Tom. ref: >List: r-sig-mixed-models >Subject: Re: [R-sig-ME] lmer() - no applicable method for 'profile' >From: Ben Bolker <bbolker () gmail ! com> >Date: 2011-01-06 15:03:52 >Message-ID: 4D25D9D8.6030402 () gmail ! com >[Download message RAW] > I believe you're stuck for the time being: profiling is not yet >implemented for GLMMs. > REML is not implemented for GLMMs either: there is some debate as to >whether a useful analogue of REML can be defined: see ><https://stat.ethz.ch/pipermail/r-sig-mixed-models/2009q1/002104.html> >for example. > I don't know of any canned approach to computing likelihood profiles >for GLMMs: there are MCMC approaches (e.g. MCMCglmm, or AD Model Builder >followed by MCMC sampling) which give you a marginal posterior >distribution ... in principle AD Model Builder can profile over the >marginal likelihood, although the last time I checked profiling didn't >actually work with random-effects models. > If you are simply trying to get confidence intervals on your >parameters, your best *simple* bet is to take the Wald test (results of >summary()). If you want a better answer than that, then I think your >choices are either an MCMC-based approach or bootstrapping (see ><http://glmm.wikidot.com/basic-glmm-simulation> to get started). > (Since you have 16 variables in the model, I hope you have at least >200-300 observations -- and that's assuming you have only main effects ...) >On 11-01-06 09:35 AM, sam steyaert wrote: > Thank you for the helping out before. I could install lme4a, and it ran > fine for all chunks of chapter 1. Anyhow, if i try with my own data, it > works, until i specify REML = FALSE in the model script, or use the update() > function. > > Then, i get the following error message (it is in fact a warning message): > > "In glmer(mymodelstructure), : > extra arguments REML are disregarded" > > I can still get the parameter estimates by calling the model name. > I would like to get the confidence intervals around the parameter estimates, > and this appears not to work. > >> prM1 = profile(Model1) > Error: is(fm@resp, "lmerResp") is not TRUE >> confint(prM1) (this function logically does not work after the former one) > Error in UseMethod("vcov") : > no applicable method for 'vcov' applied to an object of class "data.frame" > > So i guess there is something with my data structure? I use logistic > regression to model habitat use, and have 16 variables included in the > model, and one random factor (as a character). > > Does anyone has some advice? > > Thanks a lot, > > Sam > > > 2010/12/29 Douglas Bates <ba...@stat.wisc.edu> Disclaimer: click here [[alternative HTML version deleted]] ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
