On Tue, 16 Feb 1999, Peter So wrote:
> Is S-Plus capable to do reml analysis in field and agricultural
> experiments such as Genstat does?
S is a general programming language, so it is certainly capable of this,
whatever `reml analysis' is meant. Now as I understand it REML is
residual/reduced/restricted maximum likelihood, and so a method of
estimation: you also need a model to estimate for an analysis. And what is
the difference between field and agricultural experiments? I always
thought field trials were a special case of agricultural experiments.
The classical analysis of field trials would be done by the aov function in
S using an Error term, and although not usually expressed that way gives a
REML analysis. (For balanced designs the only issues are how you estimate
the variance components: for the fixed effects all reasonable methods of
estimation agree. It seems that REML was developed to extend the classical
methods to unbalanced problems.)
Two other ways to give a REML analysis in existing S software are functions
varcomp and lme, both of which can consider unbalanced designs. There is an
enhanced version of lme in beta release at
http://nlme.stat.wisc.edu/Beta
with which it is easier to analyse multistratum experiments.
All of these methods are considered in Venables & Ripley (1997, chapter
10), even illustrated on an agricultural field trial. And (my main reason
for posting a reply) the on-line statistical complements now have a chapter
showing how to use the latest lme on the examples in the book.
--
Brian D. Ripley, ripley@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 272860 (secr)
Oxford OX1 3TG, UK Fax: +44 1865 272595
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news
|