[R] "False convergence" in LME

[Andrew Robinson]

These are great tips from Spencer.

The other thing that I have found useful is to use a different
optimizing algorithm.  You can do this by:

control=lmeControl(opt = "optim")

Good luck!

Andrew


On Sat, Jun 14, 2008 at 09:45:22AM -0700, Spencer Graves wrote:
>      This is a common problem, for which solutions are poorly documented. 
> 
>           1.  Have you tried fitting simpler models, in the hopes of 
> getting something that converges without complaint, then use 'update' to 
> try more complicated models?  It sometimes works, though often not. 
> 
>           2.  Also, have you tried something like 'lme(..., 
> control=lmeControl(returnObject=TRUE))'?  If no, I suggest you try this 
> first.  With the error message you report, I would expect this to work, 
> though it may not. 
> 
>           3.  Finally, have you tried something like 'lme(..., 
> control=lmeControl(singular.ok=TRUE))' OR 'lme(..., 
> control=lmeControl(singular.ok=TRUE, returnObject=TRUE))'?  I'm not 
> sure, but I believe this may work when "returnObject=TRUE" does not. 
> 
>           4.  Have you tried the following: 
> 
>                library(FinTS)
>                package.dir('nlme')
> 
>                I tried this just now and learned that the 'nlme' 
> packages contains two non-standard subdirectories named "mlbook" and 
> "scripts".  The second contains files names 'ch01.R', 'ch02.R', etc., 
> which contain the R commands required to reproduce virtually all the 
> figures, tables and examples in Pinheiro and Bates (2000) Mixed-Effects 
> Models in S and S-PLUS (Springer).  If you have not already done so, I 
> recommend you get this book and use these script files to facilitate 
> your study of it.  Doug Bates is one of the world's leading experts in 
> mixed-effects modeling, and I have learned a lot from my study of it. 
> 
>      Hope this helps. 
>      Spencer Graves    
> 
> Rebecca Sela wrote:
> >I tried to use LME (on a fairly large dataset, so I am not including it), 
> >and I got this error message:
> >
> >Error in lme.formula(formula(paste(c(toString(TargetName), 
> >"as.factor(nodeInd)"),  : nlminb problem, convergence error code = 1
> >  message = false convergence (8)
> >
> >Is there any way to get more information or to get the potentially wrong 
> >estimates from LME?
> >
> >(Also, the page in the NLMINB documentation,  
> >http://netlib.bell-labs.com/cm/cs/cstr/153.pdf, has errors in it, which 
> >makes it harder to check on what is happening.)
> >
> >Thank you in advance!
> >
> >Rebecca
> >
> >______________________________________________
> >R-help at 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.
> >
> 
> ______________________________________________
> R-help at 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.

-- 
Andrew Robinson  
Department of Mathematics and Statistics            Tel: +61-3-8344-6410
University of Melbourne, VIC 3010 Australia         Fax: +61-3-8344-4599
http://www.ms.unimelb.edu.au/~andrewpr
http://blogs.mbs.edu/fishing-in-the-bay/