5.4.2.4 Reordering the Data Table on the Basis of an Ordination Axis
A CA result is sometimes used to reorder the data table according to the first
ordination axis. A compact form of ordered table is provided by the vegan function
vegemite(), already used in Chap. 4, which can use the information provided by
an ordination computed in vegan. One can also illustrate the ordered abundance
table as a heat map by means of function tabasco():
# Species data table ordered after the CA result
vegemite(spe, spe.ca)
# CA-ordered species table illustrated as a heat map
tabasco(spe, spe.ca)
The left-right and top-down orderings in this ordered table depend on the
(arbitrary) orientation of the ordination axes. Observe that the ordering is not
optimal since it is done only on the basis of the first CA axis. Therefore, sites 1 to
10 and 11 to 18 (separated along axis 2) and their corresponding characteristic
species are interspersed.
5.4.3 CA Using Function CA.newr()
As in the case of PCA, we propose a simple CA function: CA.newr(). Here is
how to use it on the fish data.
spe.CA.PL <- CA.newr(spe)
par(mfrow = c(1, 2))
biplot.CA(spe.CA.PL, scaling = 1, cex = 1)
biplot.CA(spe.CA.PL, scaling = 2, cex = 1)
# Ordering of the data table following the first CA axis
# The table is transposed, as in the vegemite() output
summary(spe.CA.PL)
t(spe[order(as.vector(spe.CA.PL$scaling1$sites[, 1])),
order(as.vector(spe.CA.PL$scaling1$species[, 1]))])
Hints Using $scaling2$sites and $scaling2$species (i.e. using the scaling 2
projection) would have produced the same ordered table.
Argument cex of the biplot() function is here to adapt the size of the symbols
as well as the site and species names to the plot. The default is cex = 2. Smaller
values produce smaller symbols and characters. They may be useful to draw plots
containing many sites and species.
5.4 Correspondence Analysis (CA)
181
A CA result is sometimes used to reorder the data table according to the first
ordination axis. A compact form of ordered table is provided by the vegan function
vegemite(), already used in Chap. 4, which can use the information provided by
an ordination computed in vegan. One can also illustrate the ordered abundance
table as a heat map by means of function tabasco():
# Species data table ordered after the CA result
vegemite(spe, spe.ca)
# CA-ordered species table illustrated as a heat map
tabasco(spe, spe.ca)
The left-right and top-down orderings in this ordered table depend on the
(arbitrary) orientation of the ordination axes. Observe that the ordering is not
optimal since it is done only on the basis of the first CA axis. Therefore, sites 1 to
10 and 11 to 18 (separated along axis 2) and their corresponding characteristic
species are interspersed.
5.4.3 CA Using Function CA.newr()
As in the case of PCA, we propose a simple CA function: CA.newr(). Here is
how to use it on the fish data.
spe.CA.PL <- CA.newr(spe)
par(mfrow = c(1, 2))
biplot.CA(spe.CA.PL, scaling = 1, cex = 1)
biplot.CA(spe.CA.PL, scaling = 2, cex = 1)
# Ordering of the data table following the first CA axis
# The table is transposed, as in the vegemite() output
summary(spe.CA.PL)
t(spe[order(as.vector(spe.CA.PL$scaling1$sites[, 1])),
order(as.vector(spe.CA.PL$scaling1$species[, 1]))])
Hints Using $scaling2$sites and $scaling2$species (i.e. using the scaling 2
projection) would have produced the same ordered table.
Argument cex of the biplot() function is here to adapt the size of the symbols
as well as the site and species names to the plot. The default is cex = 2. Smaller
values produce smaller symbols and characters. They may be useful to draw plots
containing many sites and species.
5.4 Correspondence Analysis (CA)
181