# Broad scale: dbMEM 1, 3, 4, 6, 7
dbmem.broad <- dbmem.red[ , 1 : 5]
# Fine scale: dbMEM 10, 11, 20
dbmem.fine <- dbmem.red[ , 6 : 8]
## 5. Mite - environment - trend - dbMEM variation partitioning
(mite.varpart # Show the symbols of the fractions and plot their values
par(mfrow = c(1,2))
showvarparts(4, bg = c("red", "blue", "yellow", "green"))
plot(mite.varpart,
digits = 2,
bg = c("red", "blue", "yellow", "green")
)
# 4. Arbitrarily split the significant dbMEM into broad and
#
fine scale
(mite.det.dbmem.fwd as.matrix(mite.dbmem),
adjR2thresh = mite.det.dbmem.R2a))
# Number of significant dbMEM
(nb.sig.dbmem <- nrow(mite.det.dbmem.fwd))
# Identify the significant dbMEM sorted in increasing order
(dbmem.sign <- sort(mite.det.dbmem.fwd$order))
# Write the significant dbMEM to a new object (reduced set)
dbmem.red <- mite.dbmem[ ,c(dbmem.sign)]
The default option of function plot.varpart() leaves the fractions with
negative R
2
adj empty in the graph. To display the negatives values, set the
argument cutoff = -Inf.
# Tests of the unique fractions [a], [b], [c] and [d]
# Fraction [a], pure environmental
anova(
rda(mite.h, env.red, cbind(mite.xy, dbmem.broad,
# Fraction [b], pure trend
anova(
rda(mite.h, mite.xy, cbind(env.red, dbmem.broad,
# Fraction [c], pure broad scale spatial
anova(rda
(mite.h, dbmem.broad, cbind(env.red, mite.xy, dbmem.fine)))
# Fraction [d], pure fine scale spatial
anova(
rda(mite.h, dbmem.fine, cbind(env.red, mite.xy,
# Tests of the unique fractions [a], [b], [c] and [d]
# Fraction [a], pure environmental
anova(
rda(mite.h, env.red, cbind(mite.xy, dbmem.broad, dbmem.fine)))
# Fraction [b], pure trend
anova(
rda(mite.h, mite.xy, cbind(env.red, dbmem.broad, dbmem.fine)))
# Fraction [c], pure broad scale spatial
anova(rda
(mite.h, dbmem.broad, cbind(env.red, mite.xy, dbmem.fine)))
# Fraction [d], pure fine scale spatial
anova(
rda(mite.h, dbmem.fine, cbind(env.red, mite.xy, dbmem.broad)))
All unique fractions are significant.
7.4 Eigenvector-Based Spatial Variables and Spatial Modelling
331