Phylogenetically aligned component analysis

It has become common in evolutionary biology to characterize phenotypes multivariately. However, visualizing macroevolutionary trends in multivariate datasets requires appropriate ordination methods. In this paper we describe phylogenetically aligned component analysis (PACA): a new ordination appro...

Full description

Saved in:
Bibliographic Details
Published in:Methods in ecology and evolution 2021-02, Vol.12 (2), p.359-372
Main Authors: Collyer, Michael L., Adams, Dean C., Blomberg, Simone
Format: Article
Language:English
Subjects:
Eigenvectors
Empirical analysis
Euclidean geometry
Hominids
Monkeys & apes
multivariate
Multivariate analysis
Ordination
Phenotypes
phylogenetic
Phylogenetics
Phylogeny
principal component
Principal components analysis
singular value decomposition
Trends
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:It has become common in evolutionary biology to characterize phenotypes multivariately. However, visualizing macroevolutionary trends in multivariate datasets requires appropriate ordination methods. In this paper we describe phylogenetically aligned component analysis (PACA): a new ordination approach that aligns phenotypic data with phylogenetic signal. Unlike phylogenetic principal component analysis (Phy‐PCA), which finds an alignment of a principal eigenvector that is independent of phylogenetic signal, PACA maximizes variation in directions that describe phylogenetic signal, while simultaneously preserving the Euclidean distances among observations in the data space. We demonstrate with simulated and empirical examples that with PACA, it is possible to visualize the trend in phylogenetic signal in multivariate data spaces, irrespective of other signals in the data. In conjunction with Phy‐PCA, one can visualize both phylogenetic signal and trends in data independent of phylogenetic signal. Phylogenetically aligned component analysis can distinguish between weak phylogenetic signals and strong signals concentrated in only a portion of all data dimensions. We provide empirical examples that emphasize the difference. Use of PACA in studies focused on phylogenetic signal should enable much more precise description of the phylogenetic signal, as a result. Overall, PACA will return a projection that shows the most phylogenetic signal in the first few components, irrespective of other signals in the data. By comparing Phy‐PCA and PACA results, one may glean the relative importance of phylogenetic and other (ecological) signals in the data.
ISSN:2041-210X
2041-210X
DOI:10.1111/2041-210X.13515