[R-sig-eco] Community composition variance partitioning?

[Alexandre Fadigas de Souza]

Hi Steve,

  Thank you for your response to my message and for the suggestion.
 
  We are also performin RDA-based variance partitioning. Reading the literature on community composition variance partition, my impression was that there is a turmoil and the field is divided into two main fields in disagreement: rda- and partial mantel-based approaches using or not pcnm as spatial descriptors (as opposed to polinomials of lat long). Simulation comparisons concluded that all approaches are subotimal and have strenghts and weakenesses. This without mentioning the danish initiative to use mixed models as a comparative means to these two approaches.

   We decided to all three: rda, mantel, and mixed model approaches, so as to be able to compare results and see if congruent patterns emerge.

   To be more specific, in the mixed model approach ordination axes (e.g., pca on hellinger-transformed species data) are used as dependent variables and explanatory environmental factors are used as independent variables. Levels of spatial cluster are included as nesting effects. Sequential model adjustment shows if space is relevent and if the environment is relevant, in which case which environmental variables are relevant are also evaluated.

  Regarding the R2 problem in the multiple regression on distance matrices, it seems that indeed the problem was that we were including variables as extra columns and not as separate matrices in the formula. With change we obtained r2 in the expected order of increase.

   What do you think of this all-inclusive approach?

   All the best,

   Alexandre

Dr. Alexandre F. Souza 
Professor Adjunto II Departamento de Botanica, Ecologia e Zoologia  Universidade Federal do Rio Grande do Norte (UFRN)  http://www.docente.ufrn.br/alexsouza  Curriculo: lattes.cnpq.br/7844758818522706

_______

Alexandre,

I'll leave it to Sarah to advise you on MRM (and I agree with Jari that
the method you're describing is not going to work). I'll just add that it
is not clear to me why the predictors (even geographic distance) have to
be treated as distances to partition the variance in composition. I'm
assuming the environmental variables were not originally in the form of
euclidean distance matrices and that the raw measurements are available?
As for the geographic distances, if you have lat and long coordinates, why
not treat both lat and long as predictors and do the necessary analyses as
partial distance-based redundancy analyses using capscale? In one analysis
the geographic predictors could be partialled out (with the result
explaining the fraction explained by the environment). In another, the
environmental predictors could be partialled out (with the result
explaining the fraction explained by the geographic distance) and in a
third both geographic and environmental predictors could be considered
with no conditioning covariates (which will give the total variance
explained by both combined).

Best
Steve