[R] Cansisc: Error in eigen(eHe, symmetric = TRUE)

[John Fox]

Dear Judith,

Michael and David's suggestion (and my earlier suggestion) that you use a
method appropriate to the distribution of your data is no doubt good advice,
but it remains to be explained why candisc() produced an error.

If you do the following, you'll see that many of your variables are
invariant within levels of sitio:

-------- snip ----------

Data <- read.csv("c:/temp/bosques1.csv", header=TRUE)
vars <- scan(what="")
biomasa altdosel altsoto cobertura riqarb elevacion temperatura 
precipitacion  Araceae Begoniaceae Bromeliaceae Clusiaceae Cyclanthaceae
Ericaceae 
Gesneriaceae Melastomataceae Orchidaceae Piperaceae Pteridophyta.taxa

by(Data[, vars], Data$sitio, function(D) apply(D, 2, sd))

-------- snip ----------

I believe that you would have discovered this had you looked at the data --
always a good idea -- prior to using candisc().

BTW, in the original script that you posted, you apparently misspelled two
of the variable names as "Cyclanthace" and "Pteridophyta", which suggests to
me that the version of the data set that you sent is different from the
version that you used.

I hope that this helps,
 John

> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-
> project.org] On Behalf Of Maria Judith Carmona H
> Sent: Thursday, July 10, 2014 11:10 AM
> To: David L Carlson
> Cc: r-help at r-project.org; John Fox; Michael Friendly
> Subject: Re: [R] Cansisc: Error in eigen(eHe, symmetric = TRUE)
> 
> Hello John, Michael and David,
> 
> I will review Vegan Package. I am very thankful with your help.
> 
> Best,
> Judith
> 
> 
> On Thu, Jul 10, 2014 at 9:04 AM, David L Carlson <dcarlson at tamu.edu>
> wrote:
> 
> > In particular, look at the vegan Vignette, "Introduction to
> Ordination in
> > vegan", particularly section 4 on constrained ordination which
> describes
> > three approaches that seem relevant to your problem.
> >
> > http://cran.r-project.org/web/packages/vegan/vignettes/intro-
> vegan.pdf
> >
> > David Carlson
> >
> > -----Original Message-----
> > From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-
> project.org]
> > On Behalf Of Michael Friendly
> > Sent: Thursday, July 10, 2014 3:22 AM
> > To: Maria Judith Carmona H
> > Cc: r-help at r-project.org; John Fox
> > Subject: Re: [R] Cansisc: Error in eigen(eHe, symmetric = TRUE)
> >
> > Maria
> >
> > The variables
> >
> >
> Araceae,Begoniaceae,Bromeliaceae,Clusiaceae,Cyclanthaceae,Ericaceae,Ges
> neriaceae,
> > Melastomataceae,Orchidaceae,Piperaceae,Pteridophyta
> > are frequencies (abundances?) of which most are 0 and this is not
> > appropriate
> > as multivariate normal data.
> >
> > There is probably a version of canonical analysis that takes such
> > variables into account,
> > but I don't know specifically.  Have you looked at the vegan package
> and
> > its references?
> >
> > HTH
> > -Michael
> >
> >
> >
> > On 10/07/2014 12:54 AM, Maria Judith Carmona H wrote:
> > > Dear John,
> > >
> > > I am including abundance values b