[R-sig-eco] Setting up dataframes for MRPP given clone library data

[michael.gruenstaeudl]

Hi all,
how does one correctly set up the input dataframes for an mrpp analysis in
VEGAN, assuming a clone sequence library is your sample data? The following
example uses the dune meadow vegetation data set to clarify the issue.

EXAMPLE:

*Input data*: 
(a) clone library sequences that have been clustered into OTUs,
(b) ecological data associated with the clone sequences (but not necessarily
with the OTUs!)

*Data matrix preparation*:
To produce the observations/sites dataframe (e.g. dune), I assign (1) clones
as observations, and (2) OTUs as sites. (Note: Since OTUs can be associated
with multiple clone sequences, but each clone sequence can only be
associated with one OTU, a dataframe set up with clones as sites and OTUs as
observations is not useable.)

To produce the environmental variables/sites dataframe (e.g. dune.env), I
assign (1) the ecological data as environmental variables, and (2) the OTUs
as sites.

*Problem*:
Given that some OTUs consist of clone sequences with different levels of the
ecological variables (e.g. "BF" for some clones and "SF" for others), I
either need to produce a combined variable level ("BFandSF"), or I split the
OTU into clones homogeneous for the respective ecological variable. Hence,
the observations/sites dataframe has now one more site.

*My thoughts*:
The first solution generates ecological variable levels that I am not
interested in.
The latter solution (done in various recent publications) might not be
statistically sound, as the within-group sites are artificially increased.

*My question*:
Is the latter approach indeed problematic (e.g. maybe causing a type I
error?), and are there strategies to circumvent the problem?

Many thanks, Michael


--
View this message in context: http://r-sig-ecology.471788.n2.nabble.com/Setting-up-dataframes-for-MRPP-given-clone-library-data-tp7335129p7335129.html
Sent from the r-sig-ecology mailing list archive at Nabble.com.