Models, Probability, and the Deep Past

Pence, Charles H. [UCL]

The case of nonidentifiable parameters in divergence time estimates in evolutionary theory offers an interesting example for the study of statistical models. Even in the limit of complete observational data, the rates and absolute times of lineage divergence still cannot be specified without statistical uncertainty. It is not true, that is, that a statistical model is only used because we are practically incapable of obtaining a sufficient quantity of data. This offers an interesting analogy with quantum mechanics, and has implications for our understanding of the role of chance in evolutionary theory.