Fig. 1: Performance of phylogenetically informed prediction against predictive equations.
From: Phylogenetically informed predictions outperform predictive equations in real and simulated data
a–d Distributions of prediction errors (actual—predicted values) from a set of 1000 simulated ultrametric trees (n = 100 tips) under three correlation coefficients (r = 0.25, 0.5, and 0.75). e–l Distributions of prediction errors from 1000 non-ultrametric trees (n = 100 tips) with low (e–h) and high (i–l) extinction rates. Boxplots elements are as follows: centre line, median; box limits, first and third quartiles; whiskers, 1.5 × inter-quartile range; points, outliers. Colours represent the three prediction methods (OLS: ordinary least squares predictive equations, green; PGLS: phylogenetic generalised least squares predictive equations, orange; phylogenetically informed predictions, blue). Tighter distributions near zero indicate more accurate predictions overall than those more skewed away from zero (measured by the variance in prediction errors, \({\sigma }^{2}\)). Phylogenetically informed predictions are more accurate than other methods. Medians and variances of distributions can be found in Supplementary Data 1.
