[R-sig-eco] adonis and distance metrics

[Etienne Laliberté]

Hi Wout,

This is always tricky. There are so many different metrics out there
that it can be really confusing to choose the "right one". Legendre and
Legendre 1998 Numerical Ecology go at great length at describing many of
them so that's a good place to start. The danger is that, not having any
idea which one to choose a priori, it can become tempting to try them
all and see which one gives you the results that you're the most happy
with (a "fishing" expedition). Because, as you seem to have noticed, the
metric can make a lot of difference.

Personally, what I would recommend, and what I've done in the last
papers where I've used such analyses, is to try a 3 different metrics
that differ in their emphasis on pure shifts in composition
(presence /absence) relative to shifts in abundance.

This is because I've been heavily influenced by Anderson et al 2006
Ecology Letters (a recommended read), who make a compelling argument
that one must be cautious at how a much a particular metric emphasizes
shifts in composition (presence/absence) vs abundance.

They propose a measure which is easy to interpret, which (confusingly
enough) they call the modified-Gower distance. What I think is neat
about it is that it is very clear what it does: with base 2, a doubling
in abundance has the same weight as a shift from 0 to 1 (i.e. from
absent to present). With base 10, then you need abundance to be 10x
greater for it to be equal to a 0->1 shift, etc. The measure is
available in vegdist:

data(varespec)
vare.dist <- vegdist(decostand(varespec, "log"), "altGower") # by
default, base 2
vare.dist <- vegdist(decostand(varespec, "log", logbase = 10),
"altGower") # base 10

But that's just me, and I'm sure others will have different opinions on
the subject. For example, it's always easier to justify the very
widely-used ones (e.g. Bray-Curtis) in a paper because every reviewer is
familiar with them...

Cheers

Etienne





Le vendredi 30 juillet 2010 à 12:39 +0200, Opdekamp Wout a écrit :
> Dear R-sig users,
> 
>  
> 
> In a full-factorial mesocosm experiment (80 mesocosms: three factors with two levels each and hence 10 replicates per treatment combination) we measured the aboveground biomass of 28 plant species that were sown in this experiment. Results are given as abundances (kg/m) in a (80 treatment-rows x 28 species-colums) dataframe. Apart from that PhospoLipicFattyAcids have also been analysed in the soil to characterise the microbial community. Results are given as peak heights in a (80 treatment-rows x 33 plfa-colums) dataframe, standardised per row.
> 
>  
> 
> I used Adonis to analyse if and how the applied factors influence these floral and microbial communities. However, I am struggling a bit with the distance metric to use as the obtained results are somewhat differed on the metric applied. Any recommendations on which one to use or where I could find a comparison of the different metrics from an ecological point of view?
> 
>  
> 
> Kind regards,
> 
> Wout
> 
>  
> 
> 
> 
> Wout Opdekamp, PhD student
> Ecosystem Management Research Group
> University of Antwerp, Belgium
> www.ua.ac.be/ecobe <BLOCKED::BLOCKED::BLOCKED::http://www.ua.ac.be/ecobe> 
> 
> P 
> 
> 
> Please consider your environmental responsibility before printing this e-mail
> 
>  
> 
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-sig-ecology mailing list
> R-sig-ecology at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-ecology


-- 
Etienne Laliberté
================================
School of Forestry
University of Canterbury
Private Bag 4800
Christchurch 8140, New Zealand
Phone: +64 3 366 7001 ext. 8365
Fax: +64 3 364 2124
www.elaliberte.info