The lack of resolution in more terminal internal nodes is common in studies using phylogenetic supertrees, but these studies also tend to have polytomous nodes basally as well [6], [11] [12].
Results for co-membership were then stable as the tree depth increased further, the main difference being more unoccupied terminal nodes and greater computational time (results not shown).
With 58, 126, and 220 parsimony informative sites for 18S, EF-1α, and COI, respectively, the nuclear genes (18S and EF-1α) provided slightly more resolution basally whereas the mitochondrial marker, COI, provided more resolution at terminal nodes (See Additional Files 2, 3, and 4).
Regardless of the type of tree depicted (Fig. 4), a rooted tree can be read as indicating the earliest ancestor at the root, from which are descended the internal nodes and, more recently, the terminal nodes.
The price paid is that stable trees that cross-validate well usually consist of no more than a few terminal nodes and are thus not very discriminating [ 24].
For the construction of the CART model, the CHAID algorithm was used; the CART architecture was 20-5-10; that is, each parent node was forced to 20 or more vectors, and each terminal node had more than 5 vectors, and the tree depth was not allowed to grow more than 10 levels.
Clustering at the terminal nodes causes more increase in NTI relative to NRI.
For example, the tree in Fig. 11 shows only a single fish, a frog, a lizard, a bird, a cat, and a human, but in actuality, these six terminal nodes together represent more than 50,000 species of living vertebrates and an untold number of ancestors.
Rather we establish a fixed tree depth, and allow the very bottom nodes, which we call terminal nodes or leaves, to have more than one graph vertex.
The dotted lines in Figure 1 show how the tree size, and hence the number of terminal nodes, would change if a more conservative alpha of 0.1 or 0.05 were selected.
