# Interpreting the medium-scaled spatial variation: regression of
# the significant spatial canonical axes on the environmental
# variables
mite.dbmemmed.ax1.env summary(mite.dbmemmed.ax1.env)
mite.dbmemmed.ax2.env summary(mite.dbmemmed.ax2.env)
The medium-scale features correspond to two types of soil coverage, to variation
in shrub density and to microtopography.
## dbMEM analysis of the mite data - fine scale
(mite.dbmem.fine anova(mite.dbmem.fine)
The analysis stops here, since the RDA is not significant.
Something has occurred here, which is often found in fine-scale dbMEM analysis. The fine-scaled dbMEM model is not significant. When significant, the finescaled dbMEM variables are mostly signatures of local spatial correlation generated
by community dynamics. This topic will be addressed later.
7.4.2.4 Hassle-Free dbMEM Analysis: Function quickMEM()
A single-step dbMEM analysis can be performed easily with our function
quickMEM(). This function only requires two arguments: a response data table
(pre-transformed if necessary) and a table containing the site geographic coordinates,
which can be one- or two-dimensional. When the default arguments are applied, the
function performs a complete dbMEM analysis: it checks whether the response data
should be detrended and does it if a significant trend is identified; it constructs the
dbMEM variables and tests the global RDA; it runs forward selection, using the
dbMEM with positive spatial correlation; it runs RDA with the retained dbMEM
variables and tests the canonical axes; it delivers the RDA results (including the set
of dbMEM variables) and plots maps of the significant canonical axes.
7.4 Eigenvector-Based Spatial Variables and Spatial Modelling
327