Re: R-beta: repeated measures

Jim Lindsey (jlindsey@luc.ac.be)
Wed, 25 Feb 1998 11:48:07 +0100 (MET)


From: Jim Lindsey <jlindsey@luc.ac.be>
Message-Id: <9802251048.AA06692@alpha.luc.ac.be>
Subject: Re: R-beta: repeated measures
To: p.dalgaard@biostat.ku.dk (Peter Dalgaard BSA)
Date: Wed, 25 Feb 1998 11:48:07 +0100 (MET)
In-Reply-To: <x2ogzw7z2t.fsf@blueberry.kubism.ku.dk> from "Peter Dalgaard BSA" at Feb 25, 98 11:34:18 am


> 
> Jim Lindsey <jlindsey@luc.ac.be> writes:
> 
> > >     Jim> R itself has nothing for repeated measures. However, I am developing a
> > >     Jim> complete set of four libraries that will handle most any repeated
> > >     Jim> measures problem, normally distributed or other. This includes two
> > >     Jim> functions in one of the libraries that will do the so-called anova
> > >     Jim> approach plus ARMA. They will do everything in my Repeated
> > >     Jim> Measurements book (OUP 1993) and much more.
> > >     Jim> I am waiting until R stabilizes a bit before releasing them but will
> > >     Jim> soon be asking for volunteers to aid in preliminary testing.
> > >     Jim> Jim
> > > 
> ...
> > > 
> > > He was also badly missing S-plus's  aov(.) function...
> > > and I explained how he could  use   anova( lm(...) ) for FIXED effects
> > > anova, and that nothing is yet available for random (or mixed) effects.
> > > 
> > > Martin.
> > > 
> > Just a warning. These things are not as simple as aov. It should still
> > be implemented! Mine are general enough to do any nonlinear model for
> > both mean and variance (e.g.PKPD), generalized linear mixed models,
> > multivariate survival with any censoring pattern, completely
> > unbalanced data, etc. Hence, with this level of generality, the
> > interface is not always that simple: lists of time-varying covariates etc.
> >   Jim
> 
> Yes. We need aov().
> 
> Also note that Bates & Pinheiro are in the process of porting lme
> (linear mixed effects models) to R. From what I have seen, this does
> at least what SAS PROC MIXED does, only better...

My libraries also do all of SAS Proc Mixed. One difference with
respect to nlme is that I have not implemented models with more than
one random effect, and certainly not for nonlinear models i.e just
random intercept with nonlinear models, but any distribution, not just
normal. I am waiting impatiently for that aspect of nlme.
  Jim
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._