# New (fictitious) objects with fish abundances
# Variables(species) must match those in the original data set in
# name, number and order
site1.new <- round(apply(spe[1:15, ], 2, mean))
site2.new <- round(apply(spe[16:29, ], 2, mean))
obj.new <- t(cbind(site1.new, site2.new))
# Hellinger transformation of the new sites
obj.new.hel <- decostand(obj.new, "hel")
# Calibration
calibrate(spe.rda.pars, obj.new.hel)
# Compare with real values at sites 7 to 9 and 22 to 24 :
env2[7:9, c(1, 9, 10)]
env2[22:24, c(1, 9, 10)]
Are the calibrated values realistic? Which ones are good, which fall far from the
true values?
Note that this is but a very small incursion into the world of environmental
reconstruction. Many more functions are available to that effect, especially in
package rioja, which is devoted to the analysis of stratigraphic data.
6.3.2.8 Variation Partitioning
A common occurrence in ecology is that one has two or more sets of explanatory
variables pertaining to different classes. In the Doubs data, we have already split the
environmental variables into a first subset of physiographic and a second subset of
chemical variables. For various reasons, one might be interested not only in a partial
analysis like the one that we conducted above, but in quantifying the variation
explained by all subsets of the variables when controlling for the effect of the
other subsets. In multivariate ecological analysis, a procedure of variation
partitioning has been proposed to that effect by Borcard et al. (1992) and improved
by the use of adjusted R
2 by Peres-Neto et al. (2006). When two explanatory data
sets are used, the total variation of Y is partitioned as in Fig. 6.5 (left). The figure also
shows the fractions resulting from partitioning by three and four sets of explanatory
variables.
# Explanation of fraction labels (two, three and four explanatory
# matrices) with optional colours
par(mfrow = c(1, 3) , mar = c(1, 1, 1, 1))
showvarparts(2, bg = c("red", "blue"))
showvarparts(3, bg = c("red", "blue", "yellow"))
showvarparts(4, bg = c("red", "blue", "yellow", "green"))
6.3 Redundancy Analysis (RDA)
233
# Variables(species) must match those in the original data set in
# name, number and order
site1.new <- round(apply(spe[1:15, ], 2, mean))
site2.new <- round(apply(spe[16:29, ], 2, mean))
obj.new <- t(cbind(site1.new, site2.new))
# Hellinger transformation of the new sites
obj.new.hel <- decostand(obj.new, "hel")
# Calibration
calibrate(spe.rda.pars, obj.new.hel)
# Compare with real values at sites 7 to 9 and 22 to 24 :
env2[7:9, c(1, 9, 10)]
env2[22:24, c(1, 9, 10)]
Are the calibrated values realistic? Which ones are good, which fall far from the
true values?
Note that this is but a very small incursion into the world of environmental
reconstruction. Many more functions are available to that effect, especially in
package rioja, which is devoted to the analysis of stratigraphic data.
6.3.2.8 Variation Partitioning
A common occurrence in ecology is that one has two or more sets of explanatory
variables pertaining to different classes. In the Doubs data, we have already split the
environmental variables into a first subset of physiographic and a second subset of
chemical variables. For various reasons, one might be interested not only in a partial
analysis like the one that we conducted above, but in quantifying the variation
explained by all subsets of the variables when controlling for the effect of the
other subsets. In multivariate ecological analysis, a procedure of variation
partitioning has been proposed to that effect by Borcard et al. (1992) and improved
by the use of adjusted R
2 by Peres-Neto et al. (2006). When two explanatory data
sets are used, the total variation of Y is partitioned as in Fig. 6.5 (left). The figure also
shows the fractions resulting from partitioning by three and four sets of explanatory
variables.
# Explanation of fraction labels (two, three and four explanatory
# matrices) with optional colours
par(mfrow = c(1, 3) , mar = c(1, 1, 1, 1))
showvarparts(2, bg = c("red", "blue"))
showvarparts(3, bg = c("red", "blue", "yellow"))
showvarparts(4, bg = c("red", "blue", "yellow", "green"))
6.3 Redundancy Analysis (RDA)
233