[R] any DCCA function in R?

Jari Oksanen jari.oksanen at oulu.fi
Tue Dec 20 15:28:02 CET 2011


Gavin Simpson <gavin.simpson <at> ucl.ac.uk> writes:

> 
> On Tue, 2011-12-20 at 09:31 +0100, Juan Santos wrote:
> > Dear members,
> > 
> > I am performing multivariate analysis on marine benthic populations 
> > using R. At first glance I found ca and VEGANO packages to be the 
> 
> That would be the `vegan` package
> 
> > suitable for the task, but neither has incorporated Detrended Canonical 
> > Correspondence Analysis (DCCA), which is just the method I want to apply 
> > on my data. I've looked for alternative packages containing the method, 
> > but my suspicion is that there is not DCCA availability for R users. 
> > Does anyone have better news for me?
> 
> Nope, sorry. I'm not aware of any such function and as I maintain the
> Environmetric Task View on CRAN I come across most ecologically-related
> packages one way or another.

I'm sure DCCA is not in vegan, and I think it is nowhere else in R.

> ii) can more easily be addressed by reducing the number of terms in the
> CCA model or a different ordination method, and i) could no doubt be
> done in a number of other ways should you really want it.
> 
Two carefully selected constraints can suffice for getting the arc effect. For
instance, the first two unconstrained CA axes or external variables similar to 
those will do the arc for you. However, usually getting an arc means that you
had two many redundant constraints.

> As far as I am aware DCCA was only ever implemented in CANOCO (possibly
> also in one of the other DOS applications from the good old days of
> quantitative ecology). Jari ported the original DECORANA code for DCA to
> vegan as it was available for us to do so. I am not aware of a
> free/unencumbered source code for DCCA, and given that and the limited
> number of applications for the method, it hasn't been something we have
> been motivated to do.

It is a non-trivial task to implement DCCA with the decorana code in vegan. 
Cajo ter Braak did a big job when he developed DCCA code for his Canoco 
(and DCCA is indeed already in the first versions of Canoco). We have no
plans of trying to start such a big job ourselves. You probably need to go to
Canoco if you *really* need DCCA.

Cheers, Jari Oksanen



More information about the R-help mailing list