[R-sig-ME] change in variance components depending on scaling of fixed effects

Martin Maechler maechler at stat.math.ethz.ch
Fri Oct 19 09:23:10 CEST 2007


>>>>> "AH" == Arild Husby <arild.husby at ed.ac.uk>
>>>>>     on Thu, 18 Oct 2007 09:53:19 +0100 writes:

    AH> Dear all,
    AH> I am trying to understand the output from a binomial lmer object and why 
    AH> the scaling of a fixed effect changes the variance components.

[....................... ]


    AH> All help is highly appreciated.

    AH> Thanks very much,
    AH> Arild Husby

    >> sessionInfo()
    AH> R version 2.5.0 (2007-04-23)
    	  	  -----
  [.......]

    AH> attached base packages:
    AH> [1] "stats"     "graphics"  "grDevices" "utils"     "datasets"  
    AH>     "methods"   "base"   
    AH> other attached packages:
    AH> lme4      Matrix     lattice
    AH> "0.99875-2" "0.99875-2"   "0.15-11"
    	 ---------

Can you  upgrade to R 2.5.1 
    (at least! --- always do upgrade from x.y.0 to x.y.1  !!! )
or better 2.6.0 and then
    upgrade.packages() ?
This will upgrade your lme4 to version  0.99875-8 (important!) 
and Matrix to 0.999375-3 {also possibly quite important}

Also, (and even alternatively to upgrading) 
if you provided a  *reproducible*  example
[e.g. you could dput(.) part of your data frame, or better
 save() it and put the file up for downloading; or use
 set.seed() and create random data which exhibits the same "feature"]
"your audience" could at least see if the current versions of
lme4 give the same results or not..

Regards,
Martin Maechler, ETH Zurich



    AH> -- 
    AH> Arild Husby
    AH> Institute of Evolutionary Biology
    AH> Room 413, Ashworth Labs,
    AH> King's Buildings,
    AH> University of Edinburgh
    AH> EH9 3JT, UK

    AH> E-mail: arild.husby at ed.ac.uk
    AH> web: http://homepages.ed.ac.uk/loeske/arild.html
    AH> Tel: +44 (0)131 650 5990
    AH> Mob: +44 (0)798 275 0668

    AH> _______________________________________________
    AH> R-sig-mixed-models at r-project.org mailing list
    AH> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models




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