[R] Design and analysis of mixture experiments

[k.jewell at campden.co.uk]

A summary, for those interested and posterity...

Thanks to Christos Hatzis who is correct, the package 'AlgDesign' (which I'd
overlooked) has gen.mixture which "Creates a candidate list of mixture
variables".
> gen.mixture(4,c("egg", "flour", "butter"))

Thanks also to a private e-mailer who suggested approaches on the lines of
> egg <- flour <- butter <- seq(0,100,length=2)
> cakemix <- expand.grid(egg=egg, flour=flour,butter=butter)
> cakemix <- cakemix[apply(cakemix,1,sum)>0,]
> cakemix <- as.data.frame(t(apply(cakemix,1,function(x) x/sum(x))))
> cakemix <- cakemix[!duplicated(cakemix),]
> cakemix

However, I was looking for something much more sophisticated, on the lines
of pages following the link in my first post
<http://www.itl.nist.gov/div898/handbook/pri/section5/pri54.htm>, at least
including extreme vertices designs.

In addition to the design, analysis of the resulting data can be (is!)
complicated by the redundancy in the design variables; one must be quite
sophisticated (by my standards!) in specifying models to be fitted to the
data and in interpreting the results. 

I was hoping that R had a package to make the whole thing easier, but I
guess I have to agree with the private e-mailer that
<QUOTE>
In fact, R is pretty ropey at this kind of classical industrial experimental
designs. You can cobble designs together, but there seems to be nothing that
straightforwardly generates things like central composites, box-behnken etc.
Obviously still waiting for someone to write the package.
</QUOTE>
I suppose this is a consequence of Rs origins in academia, and the academic
background of many of the major contributors (to whom all praise!).

I'm sure I could do this work in R (or even contribute a package!) if only I
had the necessary skills and time. Never mind, I have other tools which will
do the job. It's just a little iconoclastic finding needs that R can't meet
:-O

Regards to all.

Keith Jewell
mailto:k.jewell at campden.co.uk telephone (direct) +44 (0)1386 842055
Released by Mr. K. Jewell

> -----Original Message-----
> From: Christos Hatzis [mailto:christos.hatzis at nuverabio.com]
> Sent: 17 April 2008 17:24
> To: Jewell, Keith; r-help at r-project.org
> Subject: RE: [R] Design and analysis of mixture experiments
> 
> The place to look is the CRAN Task View 'ExperimentalDesign'.
> There are several packages there related to design and analysis of
> experiments.
> 
> The package 'AlgDesign' appears to have a function for generating
> mixture
> designs, and there might be others in other packages.
> 
> Good luck!
> 
> -Christos
> 
> > -----Original Message-----
> > From: r-help-bounces at r-project.org
> > [mailto:r-help-bounces at r-project.org] On Behalf Of
> > k.jewell at campden.co.uk
> > Sent: Thursday, April 17, 2008 11:45 AM
> > To: r-help at r-project.org
> > Subject: [R] Design and analysis of mixture experiments
> >
> > Hi,
> >
> > I'm interested in experimental design and data analysis on
> > mixtures, like cake recipes where the sum of the components
> > is fixed; e.g.
> > <http://www.itl.nist.gov/div898/handbook/pri/section5/pri54.htm>.
> >
> > I can't believe that R doesn't have facilities to design and
> > analyse such experiments, but I haven't been able to find
> > them (I have looked quite hard!). Can anyone point me in the
> > right direction?
> >
> > Thanks in advance,
> >
> > Keith Jewell
> > mailto:k.jewell at campden.co.uk telephone (direct) +44 (0)1386
> > 842055 Released by Mr. K. Jewell


_____________________________________________________________________
The information in this document and attachments is given after the exercise of all reasonable care and skill in its compilation, preparation and issue, but is provided without liability in its application or use.  It may contain privileged information that is exempt from disclosure by law and may be confidential.  If you are not the intended recipient you must not copy, distribute or take any action in reliance on it.  If you have received this document in error please notify us and delete this message from your system immediately.

Campden & Chorleywood Food Research Association Group;
Campden & Chorleywood Food Research Association (company limited by guarantee),registered number 510618);
CCFRA Group Services Ltd. (registered number 3841905); and
CCFRA Technology Ltd  (registered number 3836922),
all registered in England and Wales with the registered office at Station Road, Chipping Campden, Gloucestershire, GL55 6LD.

The Group may monitor e-mail traffic data and also the content of e-mail for the purposes of security and staff training.

Any advice given is subject to our normal terms and conditions of trading, a copy of which is available on request.

________________________________________________________________________
This e-mail has been scanned for all viruses by MessageL...{{dropped:2}}