Classic · The paper that put trees in every regression
Felsenstein counts the data points, and there are fewer than you think
Felsenstein J. Phylogenies and the comparative method. The American Naturalist 125:1–15, 1985. doi:10.1086/284325
Correlating two traits across species had long been treated as ordinary regression with one observation per species. Felsenstein showed with a simple branching simulation that this is wrong: closely related species resemble each other because they share history, so the observations are not independent, and the resulting confidence intervals can be badly overconfident. His fix was independent contrasts: under Brownian motion on a known tree, differences between sister nodes give n minus 1 quantities that are independent with calculable variances. Everything downstream, phylogenetic generalized least squares, Ornstein-Uhlenbeck models, phylogenetic mixed models, grew from that argument.