Figure 3 - PCA of harmonic coefficients, black outlines visualise
outlines of representative pedipalps, grey outline are hypothetical
morphs (calculated at the 20th harmonic) from extreme
position on the PC axes, convex hulls represent the intraspecific
variation within species. Exo = Phrynichus exophalmus , dia =Damon diadema , med = Damon medius , lon =Heterophrynus longicornis , whi = Phrynus whitei , cor =Acanthophrynus coronatus , ger = Phyrnus gervasii , vir =Paraphrynus viridiceps , lop = Phrynus longipies , azt =Paraphrynus aztecus , wil = Paraphrynus williamsi
Species clustering using k-means and MANOVA methods provided markedly
different results. In the femur, unsupervised k-means clustering could
only classify Ph. exophalmus with 100% accuracy, placing all
specimens of this species within the same cluster alongside no
additional species. K-means also placed the two Damon species
together within a discrete genus cluster which contained no other
species, but was unable to differentiate between the two species.
K-means performed poorly at correctly identifying all other species (see
Table 1). In contrast, MANOVA found statistically significant
differences in the pairwise comparisons between most species in the
femur (see Table 2). However, like k-means clustering, MANOVA was unable
to delimitate between species within the genera Paraphrynus and