Hi R users,
I have two data matrix, one with community data and another with
environmental data. Prior to preform the CCA, I have used PCA to select some
environmental variables and to avoid redundance information. The result is
that I have 4 environmental variables and my community data matrix where,
following bibliography, I have eliminated rare species.
All variables were log-transformed (x+1) befor the analysis was carried out
since they did not fit a normal distribution and to improve linearity.
I performed cca analysis as follows in R:
CCA1<-cca(spp.data~VAR1+VAR2+VAR3+VAR4, data=DATA)
and I used anova(CCA1, perm.max=499) to test the significance by means of
Monte Carlo permutations under full model.
The model was p<0.05 and the result of the plot was "good for my eyes",
however, when I did summary(CCA1), the first two axis accounted 0.04 CCA1
and similar in CCA2....then the variation explained by each axis was small.
On the other hand, when I performed CCA in CANOCO, without selecting the
option Log-transforming data matrix and without downweighting rare species,
the results were the opposite from the CCA performed in R. The
axes accounted high percentage of variation, the model was also significant,
but the plot had little sense.
Thank you very much!
[[alternative HTML version deleted]]