Consistency() calculates the consistency "index" and retention index
Farris1989TreeSearch
for each character in a dataset, given a bifurcating tree.
Although there is not a straightforward interpretation of these indices,
they are sometimes taken as an indicator of the fit of a character to a
tree.
Values correlate with the number of species sampled and the
distribution of taxa between character states, so are not strictly comparable
between characters in which these factors differ.
Arguments
- dataset
A phylogenetic data matrix of phangorn class
phyDat, whose names correspond to the labels of any accompanying tree.- tree
A tree of class
phylo.- compress
Logical specifying whether to retain the compression of a
phyDatobject or to return a vector specifying to each individual character, decompressed using the dataset'sindexattribute.
Value
Consistency() returns a matrix with named columns specifying the
consistency index (ci),
retention index (ri), and
rescaled consistency index (rc).
Details
The consistency "index" Kluge1969TreeSearch is defined as the number of steps observed in the most parsimonious mapping of a character to a tree, divided by the number of steps observed on the shortest possible tree for that character. A value of one indicates that a character's fit to the tree is optimal. Note that as the possible values of the consistency index do not range from zero to one, it is not an index in the mathematical sense of the term.
The maximum length of a character (see MaximumLength()) is the
number of steps in a parsimonious reconstruction on the longest possible tree
for a character.
The retention index is the maximum length of a character minus the number
of steps observed on a given tree; divided by the maximum length minus the
minimum length. It is interpreted as the ratio between the observed
homoplasy, and the maximum observed homoplasy, and scales from zero
(worst fit that can be reconstructed under parsimony) to one (perfect fit).
The rescaled consistency index is the product of the consistency and retention indices; it rescales the consistency index such that its range of possible values runs from zero (least consistent) to one (perfectly consistent).
The lengths of characters including inapplicable tokens are calculated
following Brazeau2019;textualTreeSearch, matching their
default treatment in TreeLength().
Examples
data(inapplicable.datasets)
dataset <- inapplicable.phyData[[4]]
head(Consistency(dataset, TreeTools::NJTree(dataset)))
#> ci ri rc
#> [1,] 0.2500000 0.6250000 0.1562500
#> [2,] 0.3333333 0.3333333 0.1111111
#> [3,] 0.3333333 0.6000000 0.2000000
#> [4,] 0.2500000 0.2500000 0.0625000
#> [5,] 0.5000000 0.8333333 0.4166667
#> [6,] 0.2500000 0.2500000 0.0625000