Skip to content

lme syntax for P&B examples

2 messages · Doran, Harold, Martin Maechler

Doran, Harold
# 
Paul:

It is a little difficult to understand what you are trying to translate
since you do not show what the model would look like using lme. If you
show lme, then it is easy to translate into lmer syntax.

A few thoughts, first, use lmer in the Matrix package and not in lme4.
Second, see the Bates article in R news at the link below for dealing
with nesting structures. Last, a colleague and I have a paper in press
showing how to fit models using lme which we submitted a year or so ago.
Since lme has evolved to lmer, we created an appendix that translates
all of our lme models to the lmer syntax so readers can see
equivalences. I am happy to send this to you (or others) upon request.

http://cran.r-project.org/doc/Rnews/Rnews_2005-1.pdf

Harold


 

-----Original Message-----
From: r-help-bounces at stat.math.ethz.ch
[mailto:r-help-bounces at stat.math.ethz.ch] On Behalf Of Paul Cossens
Sent: Wednesday, February 08, 2006 12:08 AM
To: r-help at stat.math.ethz.ch
Subject: [R] lme syntax for P&B examples

Hi helpeRs,
 
I've been working through some examples in Pinhiero & Bates( 2000)
trying to understand how to translate to the new Lme4 syntax but without
much luck.
Below is what I think I should do, but either the answers don't come out
the same or I get errors. 
In the Oxide problems I'm particularly interested in obtaining the
levels coeficients but this options no longer seems to be available in
lme4. How can levels infor be obtained in lme4?
 
If someone can recreate the examples below in lme4 syntax so I can
follow what is happening in the text I'd be grateful. 
 
Cheers
 
Paul Cossens
 
 
#Pixel
# P&B(2000) p40-45
 
Pixel<-read.csv("Pixel.csv",header=TRUE);
Pixel$Side<-as.factor(Pixel$Side)
Pixel$Dog<-as.factor(Pixel$Dog)
 
(fm1Pixel <- lmer(pixel ~ day + I(day^2) +(day|Dog)+(1|Side), data =
Pixel))
(fm2Pixel <- lmer(pixel ~ day + I(day^2) +(day|Dog), data = Pixel))
(fm3Pixel <- lmer(pixel ~ day + I(day^2) +(1|Dog:Side), data = Pixel))
or should I do it this way?
Pixel$DS<-with(Pixel,Dog:Side)[drop=TRUE]
(fm3Pixel <- lmer(pixel ~ day + I(day^2) +(1|DS), data = Pixel))
 
(fm4Pixel <- lmer(pixel ~ day + I(day^2) +Side , data = Pixel))
 

#Oxide
# P&B(2000) p167-170
 
Oxide<-read.csv("Oxide.csv",header=TRUE);
Oxide$Source<-as.factor(Oxide$Source)
Oxide$Lot<-as.factor(Oxide$Lot)
Oxide$Wafer<-as.factor(Oxide$Wafer)
Oxide$Site<-as.factor(Oxide$Site)
fm1Oxide<-lmer(Thickness~ (1|Lot)+(1|Lot:Wafer),data=Oxide) )
(fm2Oxide<-lmer(Thickness~ (1|Lot),data=Oxide) )
coef(fm1Oxide)


______________________________________________
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
1 day later
Martin Maechler
# 
HaroldD> ................

    HaroldD> A few thoughts, first, use lmer in the Matrix
    HaroldD> package and not in lme4. 

Well, that's potentially misleading advice:  If both lme4 and
Matrix are (more or less) current, it is actually better to call
library(lme4) if you want to use  lmer() .

A more extensive advice + explanation :

  - "lme4" is the package you should use when using lmer().
    Using "Matrix" *instead* works as well *currently*
    (and yes, *currently* lmer is in the Matrix namespace).

  - At the moment, "lme4" is an almost empty wrapper around using
    package "Matrix", since the C code for lmer() needs to be
    able to call C code inside "Matrix" and that is not yet
    portably possible.
    For that reason, everything for lmer() is currently inside
    "Matrix", but that should change when R will have the
    infrastructure for R packages to export their C API such that
    other R packages can reliably use that C API.


    HaroldD> ................

    HaroldD> Last, a colleague and I have a paper in press
    HaroldD> showing how to fit models using lme which we
    HaroldD> submitted a year or so ago.

    HaroldD> Since lme has evolved to lmer, we created an
    HaroldD> appendix that translates all of our lme models to
    HaroldD> the lmer syntax so readers can see equivalences. I
    HaroldD> am happy to send this to you (or others) upon request.

That sounds quite interesting; I'll be glad to receive it as
well.

Martin Maechler, ETH Zurich


    HaroldD> http://cran.r-project.org/doc/Rnews/Rnews_2005-1.pdf

    HaroldD> Harold