[R-sig-eco] when to use PCNM and other questions about varpart in Vegan

[ckellogg]

Hi All,
I am planning to use varpart in {Vegan} to partition the variance in my
bacterial community composition between 14 sites in the Arctic Ocean between
space and environmental variables.  I am noticing that in many instances, a
PCNM is employed to remove spatial autocorrelation.  But, I think I am not
quite understanding what this does exactly and how I use this prior to doing
the variance partitioning.  Maybe i can try and explain what I am trying to
do.  Basically, I have 14 sites and at each site 3 samples were taken at 2
to 3 depths to for bacterial community composition analysis.  I also took
samples for environmental data at each depth.  I have distance calculated
between sites over water (not over land or 'as the crow flies'), referenced
to a point on land (i also have the latitude and longitude of each site; but
i wasn't sure that these coordinates themselves were the best measure of
distance between sites).  Ultimately, I am trying to figure out how much of
the variation in the bacterial community between sites is attributable to
environmental variability as well as to geographic distance between the
sites.  Surely, the environmental variables also have some spatial component
to them as well.  So, my plan was to use varpart as follows: 

BACvsEnvvsGeo<-varpart(ANbac,ANenv.z,ANdist_MR.z) 
plot(BACvsEnvvsGeo) 

...where ANbac = bacterial community matrix (site x species), ANenv.z =
log-transformed and standardized environmental data (site x env parameters),
ANdist_MR.z = log-transformed and standardized distance between cites
referenced to a point land to the west of all sites, site x distance (1
column))

I am planning to do this in conjunction with an RDA for space and
environmental data separately.  This would allow me to calculate how much
variation is do to environmental variation alone, space alone, and
environment+space.  So, my questions are: (1) do I need to do a PCNM
beforehand.  And do I do this just on the distance between sites data. Or do
this also on the environmental data.  (2) And ultimately, what does doing
the PCNM get me.  (3) Also, I am wondering if I really should use Lat/Long
data instead of distance between sites.  I am not really sure it matters,
but I was not sure if it matter for the analysis.  (4) And, my bacterial
community data is standardized to total abundance at each size (so, percent
abundance), and I am wondering if it is necessary to do further
transformations of the data.  I see that this standardization is an option
in the decostand algorithm, so I am think that it is okay to not further
transform my data. (5) Alternatively, can varpart be used with db-RDA, so
that I can put in my distance matrix instead of the transformed species
data.  If so, how can I do this?

Sorry for the onslaught of questions.  There are not really any people in my
department that I can ask these questions too...

Thank you very VERY much for any advice.  I am really having a hard time
deciding/figuring out which is the best way to proceed.    

Colleen

----
Colleen Kellogg
PhD Candidate
School of Oceanography
University of Washington
Box 357940
Seattle, WA 98195

--
View this message in context: http://r-sig-ecology.471788.n2.nabble.com/when-to-use-PCNM-and-other-questions-about-varpart-in-Vegan-tp6295323p6295323.html
Sent from the r-sig-ecology mailing list archive at Nabble.com.