[R] [OT] Appropriate test?

[ripley@stats.ox.ac.uk]

On Wed, 12 Mar 2003, Spencer Graves wrote:

> If you want to use nlme, am I correct is asserting that the best 
> reference is
> 
> Pinhiero and Bates (2000) Mixed Effects Models in S and S-Plus (NY: 
> Springer) ?

Yes.

There are several other ways to do this though, and I don't think I would 
start with nlme.  The book by Diggle, Liang and Zeger is better at
showing the range of possibilities.  Note that you are rather short on 
subjects: with 8 measures on only 8 subjects (in the smaller group)
it is going to be hard to generalize to a population.  So your best hope 
is if the pattern of responses is simple (e.g. linear).

The `classical' approach is that you have an 8-way response for each 
subject, in which case you can use MANOVA (see ?summary.manova in R) to 
test the difference between groups.

Probably the most fruitful appoach is to plot all 20 responses against
time (by a lattice plot?) and then see if a simple model will summarize
all the curves.  You could then use (n)lme, but for the purposes of
testing the difference between the groups I would start by fitting the
model to each subject, and use the parameter(s) of each model as input to
MANOVA (or a multivariate T test).  The real advantage of (n)lme would
come if you have different measurement times for each subject: there are
examples of repeated measures studies of that form in MASS4 (Venables &
Ripley, 2002, see the R FAQ).

You could use aov() with an Error term.

You could use GEE.

Jim Lindsey's packages had various functions for repeated measures last 
time I looked.

> 
> Spencer Graves
> 
> kjetil brinchmann halvorsen wrote:
> > On 12 Mar 2003 at 2:37, Blaise TRAMIER wrote:
> > 
> > You should have a look at the package nlme, installed with your 
> > distribution of R. Look at the help for the function 
> > lme (linear mixed models). You have not described your data
> > sufficiently to say much more, but you have biological measurements 
> > changing through time, if that is in some non-linear fashion
> > you could have use of the function nlme (non linear mixed models). 
> > 
> > If this is'nt enough to get you going, come back with more
> > details of your data AND questions to answer.
> > 
> > Kjetil Halvorsen
> > 
> > 
> >>Hi,
> >>	I'm having some problem with a dataset and I don't really know how to 
> >>analyse it.
> >>
> >>I have 20 subjects in two groups of treatment (8 an 12 subjects).
> >>Biological measure have been recorded at different time, from t0 (before 
> >>the treatment) to t7 (3 days after). The time elapsed between each 
> >>measure is not constant.
> >>
> >>What is the most appropriate test to show a difference between the 2 
> >>treatements?
> >>
> >>I thought that an anova for repeated measure could do the trick, but I 
> >>didn't really find how to do it with R.
> >>
> >>I sorry for being OT but I didn't really know where to ask this 
> >>question. If you could redirect me on a appropriate forum or mailing 
> >>list to have that sort of help, I would appreciate a lot.
> >>
> >>Thanks in advance.
> >>
> >>-- 
> >>Blaise TRAMIER
> >>
> >>______________________________________________
> >>R-help at stat.math.ethz.ch mailing list
> >>https://www.stat.math.ethz.ch/mailman/listinfo/r-help
> > 
> > 
> > ______________________________________________
> > R-help at stat.math.ethz.ch mailing list
> > https://www.stat.math.ethz.ch/mailman/listinfo/r-help
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-help
> 

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595