Supporting Information (Figs S14 - S
Regression Tree for Habitat Dimension
bdbm>=0.4855
|
pelv.l< 0.0855
-1.615
n=6
bdbm>=0.425
-0.4622
n=20
0.1932
n=17
1.204
n=13
Figure S14: Regression tree for the habitat dimension. Regression tree was constructed
using the package RPART in R. Full tree was grown using the function rpart() with a
minimum split condition of n=2 and subsequently pruned using the 1-SE criterion and the
function prune(). The response variables used were PC1 and PC2 from the principal
components analysis of the habitat dimension (Table S1) and the explanatory variables
used were body depth, body width, body depth below midline (bdbm), mouth position,
pectoral fin length, pectoral fin height, caudal fin length, caudal fin height, and pelvic fin
length (pelv.l), with fin lengths expressed as ratios of standard length. Species move to
the left when the stated condition is true. Species groupings at terminal nodes were
subsequently used as the categories for each dimension while building the niche
classification scheme in Box 2.
Regression Tree for Life History Dimension
season< 3.5
|
parent< 1.5
fecundity>=931.5
-1.056
n=12
parent< 3
0.0481
n=7
0.4199
n=14
1.222
n=10
-0.4432
n=13
Figure S15: Regression tree for the life history dimension. Regression tree was
constructed using the package RPART in R. Full tree was grown using the function
rpart() with a minimum split condition of n=2 and subsequently pruned using the 1-SE
criterion and the function prune(). Species move to the left when the stated condition is
true. The response variables used were PC1 and PC2 from the principal components
analysis of the life history dimension (Table S2) and the explanatory variables used were
generation time, reproductive season (season), reproductive bouts, fecundity, egg
diameter, parental care (parent), and standard length. Species groupings at terminal
nodes were subsequently used as the categories for each dimension while building the
niche classification scheme in Box 2.
Regression Tree for Trophic Dimension
fish>=0.0042
|
macrophytes>=0.03265
decapod>=0.1981
fish>=0.2732
-2.267
n=2
-1.057
n=10
-0.0873
n=16
-0.3076
n=5
microorganisms< 0.08285
0.581
n=18
1.516
n=5
Figure S16: Regression tree for the trophic dimension. Regression tree was constructed
using the package RPART in R. Full tree was grown using the function rpart() with a
minimum split condition of n=2 and subsequently pruned using the 1-SE criterion and the
function prune(). Species move to the left when the stated condition is true. The response
variables used were PC1 and PC2 from the principal components analysis of the trophic
dimension (Table S3) and the explanatory variables used were volumetric proportions of
detritus, algae, macrophytes, microorganisms, worms/molluscs, microcrustacea, decapod
crustaceans (decapod), aquatic insects, terrestrial insects, and fish. Species groupings at
terminal nodes were subsequently used as the categories for each dimension while
building the niche classification scheme in Box 2.
Regression Tree for Defense Dimension
armor< 1.5
|
-0.3143
n=45
1.286
n=11
Figure S17: Regression tree for the defense dimension. Regression tree was constructed
using the package RPART in R. Full tree was grown using the function rpart() with a
minimum split condition of n=2 and subsequently pruned using the 1-SE criterion and the
function prune(). Species move to the left when the stated condition is true. The response
variables used were PC1 and PC2 from the principal components analysis of the defense
dimension (Table S4) and the explanatory variables used were spines, venom, armor,
aggression, crypsis, speed, and body diameter. Species groupings at terminal nodes were
subsequently used as the categories for each dimension while building the niche
classification scheme in Box 2.
Regression Tree for Metabolic Dimension
respiration< 1.5
|
activity>=2.5
fat>=3.25
-0.6001
n=20
fat>=2.75
0.2371
n=7
-0.205
n=15
activity>=1.5
0.4273
n=7
1.306
n=6
2.587
n=1
Figure S18: Regression tree for the metabolic dimension. Regression tree was
constructed using the package RPART in R. Full tree was grown using the function
rpart() with a minimum split condition of n=2 and subsequently pruned using the 1-SE
criterion and the function prune(). Species move to the left when the stated condition is
true. The response variables used were PC1 and PC2 from the principal components
analysis of the metabolic dimension (Table S5) and the explanatory variables used were
activity level (activity), hypoxia tolerance, visceral fat storage (fat), and special accessory
respiration (respiration). Species groupings at terminal nodes were subsequently used as
the categories for each dimension while building the niche classification scheme in Box 2.