[R-sig-ME] About lmer() function in package lme4

[Martin Maechler]

>>>>> "ET" == Edgar Torres <Edgar.Torres at ifremer.fr>
>>>>>     on Thu, 01 Apr 2010 17:56:22 +0200 writes:

    ET> Hello,

    ET> I have had problems with the function lmer in the package lme4, I'm  
    ET> trying to run a GLMM with a poisson distribution. But I receive the  
    ET> warning message

    ET> Warning message: In mer_finalize(ans) : iteration limit reached  
    ET> without convergence (9)

    ET> I tried to put more iteration with control=list(maxIter=500), but  
    ET> actually the program justo run 300 times, as the default.

    ET> the code that I wrote is this:

    ET> GLMM1 <- lmer(OPTOT ~ fYEAR + (fYEAR | BOAT), data=DATA1,  
    ET> family=poisson, na.action=na.omit,  
    ET> control=list(maxIter=500,msVerbose=T), verbose=T)

Coincidence ?

Just Tuesday and yesterday, Doug Bates and were prompted to fix
this thanks to an (R-forge) bug report + e-mail from Ben Bolker.
Doug released the new version of lme4 to CRAN's incoming,
but because of Easter holidays, those packages are not yet
available for installation.

However, if you install lme4 from R-forge,
you should get a version that fixes your problem.

After you do
  >  install.packages("lme4", repos="http://r-forge.r-project.org")

  >  packageDescription("lme4")

should contain
       Repository/R-Forge/Revision: 716
(or a higher number than 716).
In that case, you should be fine.

Can you confirm?

Regards,
Martin Maechler


    ET> I am not really sure that the way that I wrote the control statement  
    ET> is the right one.

    ET> Thanks a lot

    ET> Edgar TORRES
    ET> Phd student in Ecosystems
    ET> Centre de Recherche Halieutique  (IRD)
    ET> Avenue Jean Monnet BP 171
    ET> 34203 Sète cedex France