作者: José Alexandre Felizola Diniz Filho , Fabricio Villalobos , Luis Mauricio Bini
DOI: 10.1590/S1415-475738320140391
关键词: Multidimensional scaling 、 Regression 、 Biology 、 Pattern recognition 、 Phylogenetic comparative methods 、 Artificial intelligence 、 Eigenfunction 、 Phylogenetic tree 、 Autocorrelation 、 Phylogenetics 、 Bioinformatics 、 Pairwise comparison
摘要: Eigenfunction analyses have been widely used to model patterns of autocorrelation in time, space and phylogeny. In a phylogenetic context, Diniz-Filho et al. (1998) proposed what they called Phylogenetic Eigenvector Regression (PVR), which pairwise distances among species are submitted Principal Coordinate Analysis, eigenvectors then as explanatory variables regression, correlation or ANOVAs. More recently, new approach Mapping (PEM) was proposed, with the main advantage explicitly incorporating model-based warping distance an Ornstein-Uhlenbeck (O-U) process is fitted data before eigenvector extraction. Here we compared PVR PEM respect estimated signal, correlated evolution under alternative evolutionary models imputation, using simulated data. Despite similarity between two approaches, has slightly higher prediction ability more general than original PVR. Even so, conceptual sense, may provide technique best both worlds, combining flexibility data-driven empirical eigenfunction sounding insights provided by well known comparative analyses.