tree.build {BAT}

Build functional tree.

Description

Builds a functional tree from trait or distance data.

Usage

tree.build(
  trait,
  distance = "gower",
  func = "nj",
  fs = 0,
  convert = NULL,
  weight = NULL,
  root = NULL
)

Arguments

trait

A species x traits matrix or data.frame or, alternatively, a dist object.

distance

One of "gower" or "euclidean". Not used if trait is already a dist object.

func

One of "upgma", "mst", "nj", "bionj" or "best".

fs

Only used for func = "nj" OR "bionj". Argument s of the agglomerative criterion: it is coerced as an integer and must at least equal to one.

convert

A vector of column numbers, usually categorical variables, to be converted to dummy variables. Not used if trait is already a dist object.

weight

A vector of column numbers with weights for each variable. Its length must be equal to the number of columns in trait. Not used if trait is already a dist object.

root

A numeric or character specifying the functional outgroup to root the tree.

Details

The tree will be built using one of four algorithms after traits are dummyfied (if needed) and standardized (always): If func = "upgma" uses average linkage clustering (UPGMA, Cardoso et al. 2014). If func = "mst" uses minimum spanning trees, equivalent to single linkage clustering (Gower & Ross 1969). If func = "nj" uses the original neighbor-joining algorithm of Saitou & Nei (1987). If func = "bionj" uses the modified neighbor-joining algorithm of Gascuel (1997). Any of the neighbor-joining options is usually preferred as they keep distances between species better than UPGMA or MST (Cardoso et al. subm.). If func = "best", chooses the best of the options above based on maximum tree.quality values. If NJ trees are built, the root will be set at the node closest to the midpoint between the two most dissimilar species in the tree or, if root not NULL, at the node provided in parameter root (Podani et al. 2000). Gower distance (Pavoine et al. 2009) allows continuous, ordinal, categorical or binary variables, with possible weighting. NAs are allowed as long as each pair of species has at least one trait value in common. For fs > 0 even if this condition is not met the Q* criterion by Criscuolo & Gascuel (2008) is used to fill missing data. If convert is given the algorithm will convert these column numbers to dummy variables. Otherwise it will convert all columns with factors or characters as values.

Value

A phylo object representing a functional tree.

References

Cardoso et al. (2014) Partitioning taxon, phylogenetic and functional beta diversity into replacement and richness difference components. Journal of Biogeography, 41: 749-761.

Cardoso et al. (subm.) Using neighbor-joining trees for functional diversity analyses.

Criscuolo & Gascuel (2008) Fast NJ-like algorithms to deal with incomplete distance matrices. BMC Bioinformatics, 9: 166.

Gascuel (1997) BIONJ: an improved version of the NJ algorithm based on a simple model of sequence data. Molecular Biology and Evolution, 14: 685–695.

Gower & Ross (1969) Minimum spanning trees and single linkage cluster analysis. Journal of the Royal Statistical Society, 18: 54-64.

Pavoine et al. (2009) On the challenge of treating various types of variables: application for improving the measurement of functional diversity. Oikos, 118: 391-402.

Podani et al. (2000) Additive trees in the analysis of community data. Community Ecology, 1, 33–41.

Saitou & Nei (1987) The neighbor-joining method: a new method for reconstructing phylogenetic trees. Molecular Biology and Evolution, 4, 406–425.

Examples

trait = data.frame(body = c(NA,2,3,4,4), beak = c(1,1,1,1,2))
plot(tree.build(trait))
plot(tree.build(trait, func = "bionj", fs = 1, weight = c(1, 0)), "u")
plot(tree.build(trait, func = "best", root = 4))