relative compositional dissimilarity among sites against a continuous variable

Thanks Jari, so in essence I would be doing something like a PERMANOVA.

-----Original Message-----
From: Jari Oksanen [mailto:jari.oksanen at oulu.fi]
Sent: Wednesday, November 24, 2010 5:35 PM
To: Pekin, Burak K; 'Andy Rominger'
Cc: 'r-sig-ecology at r-project.org'
Subject: Re: [R-sig-eco] relative compositional dissimilarity among sites
against a continuous variable

On 24/11/10 23:48 PM, "Pekin, Burak K" <bpekin at purdue.edu> wrote:

Thanks Andy, that's a simple enough calculation. So the larger the
average dissimilarity, the more the site is dissimilar in its
composition to other sites. I could then do a regression between the
average dissimilarity of sites and other site factors and say
something like 'compositional dissimilarity increases as factor X increases'.

Burak,

If this is really what you need, remember that diagonal elements of the
dissimilarity matrix are zeros (dissimilarity between the object and itself).
This means that the row or column mean of the dissimilarity to
*other* sites is underestimated. You should either make the diagonal to NA:

d <- as.matrix(vegdist(x))
diag(d) <- NA
rowMeans(d, na.rm = TRUE)

or adjust the means for one zero element:

d <- as.matrix(vegdist(x))
rowMeans(d)*(nrow(d)/(nrow(d)-1))

This does not influence the ranks, but it doesn't hurt if your numbers are
correct.

Is there a method that regresses the relative compositional
dissimilarity among sites against a continuous variable?

The mean dissimilarity is found for your observations, so it is of the same
length as your data vectors. You can use functions like lm(), glm(), nls(),
gam() or some hundred other alternatives in packages to analyse the results.

Cheers, Jari Oksanen
 

-Burak

From: Andy Rominger [mailto:ajrominger at gmail.com]
Sent: Tuesday, November 23, 2010 4:11 PM
To: Steve_Friedman at nps.gov
Cc: Pekin, Burak K; r-sig-ecology-bounces at r-project.org;
r-sig-ecology at r-project.org
Subject: Re: [R-sig-eco] Dissimilarity ranking

or perhaps could you use something as simple as the mean
dissimilarity?  i'm thinking something like this:

library(vegan)

##    some data to play with, 20 sites by 30 spp
data(dune)

##    calculate distance, making it a matrix
dune.dist <- as.matrix(vegdist(dune,method=
"bray"))

##    take average distance
dune.avg <- apply(dune.dist,1,mean)
dune.avg

not sure what the gurus of dissimilarity analysis might think of this.
good luck,
andy

On Tue, Nov 23, 2010 at 5:43 PM,
<Steve_Friedman at nps.gov<mailto:Steve_Friedman at nps.gov>> wrote:
Burak

You should explore the use of non-metric multidimensional scaling.

?cmdscale

Steve Friedman Ph. D.
Ecologist  / Spatial Statistical Analyst Everglades and Dry Tortugas
National Park
950 N Krome Ave (3rd Floor)
Homestead, Florida 33034

Steve_Friedman at nps.gov<mailto:Steve_Friedman at nps.gov>
Office (305) 224 - 4282
Fax     (305) 224 - 4147



            "Pekin, Burak K"
            <bpekin at purdue.ed
            u>                                                         To
            Sent by:
"'r-sig-ecology at r-project.org<mailto:r-sig-ecology at r-project.org>'"
            r-sig-ecology-bou
<r-sig-ecology at r-project.org<mailto:r-sig-ecology at r-project.org>>
            nces at r-project.or<mailto:nces at r-project.or>
cc
            g
                                                                  Subject
                                      [R-sig-eco] Dissimilarity ranking
            11/23/2010 03:09
            PM








Hello, I want to rank the dissimilarity of sites based on their
species composition. For example, I would like to be able to say that
site A is less similar in composition to the other sites than site B
is similar to the other sites. I could do a cluster analysis and look
at which sites are less closely clustered.

It would be even better if I could come up with a quantitative scale
rather than a relative ranking that would give a value for each site
based on its relative dissimilarity to the rest of the sites. So site
A might receive a
90 out of 100, whereas site B and C might receive a 60 and a 50
indicating the rank as well as 'relative quantity' of dissimilarity for each
site.

Thanks,
Burak

--------------------------

Burak K. Pekin, PhD
Postdoctoral Research Associate
Purdue University


            [[alternative HTML version deleted]]