[R-sig-eco] Quantifying functional trait diversity through Gower distance and pcoa

[Jari Oksanen]

On 16/11/2010 17:08, "chris mcowen" <chrismcowen at gmail.com> wrote:

> Dear List,
> 
> I am relatively new to this area of ecology and R, coming from a
> phylogenetic background. I am looking to quantify functional trait
> diversity in in order to plot global functional trait diversity
> hotspots. My trait data are categorical and all indicies of functional
> diversity require continuous data.
> 
> My approach was to generate a distance matrix based on the Gower
> method - i have done this with vegdist- and then i was planning on
> running a pcoa on these to generate the continuous data to use in the
> calculation of functional diversity indicies.
> 
> I have two questions:-
> 
> First, does this make sense, i appreciate i will loose information this way.
> 
> Second, I have looked at vegan but can not see a way to run a pcoa on
> the output from a vegdist analysis, is this possible and once again
> does this make sense?
>

Chris,

PCoA is standard R: use function cmdscale(). There is no need to use any
special packages, although PCoA is also in many special packages (including
ade4 and even vegan). There is an unfortunate confusion with names. Gower
suggested a new name to metric or classic scaling, but not many people
outside ecology followed his suggestion. Now we are in trouble in searching
the method. It is usually called metric or classic multidimensional scaling,
hence the name cmdscale in base R (or 'stats' for pedantic).  Perhaps
somebody could suggest Ripley that he adds "principal coordinates analysis"
as a subject in the cmdscale help page to make easier to find this function.

Package FD indeed stands for *F*unctional *D*iversity, but scattered
functions can be found elsewhere, also in vegan (treedive).

Cheers, Jari Oksanen