A Probability Model for Inferring Evolutionary Trees

Estimation of evolutionary trees should be treated as a problem in statistical inference, but such treatment requires the explicit formulation of a stochastic model of the evolutionary process. Because an evolutionary inference procedure is likely to be put to such uses as deciding the issue of whether rates of evolution are homogeneous, the stochastic model underlying the inference procedure should not assume homogeneity over time of the evolutionary process, and in fact, should make only the weakest evolutionary assumptions necessary. Such a model is constructed, and it is shown that most parsimonious trees are maximum-likelihood estimated evolutionary trees under the stochastic model. Similarity clustered phenograms appear not to be well justified as statistical estimates of evolutionary trees, even when homogeneity of evolutionary rates is assumed.