Title: Information in uncertainty: Likelihood inference in latent variable problems.
Abstract: MCMC or other Monte Carlo approaches enable estimation of likelihoods of models of complex systems involving latent variables. Such models arise in a variety of areas of inference in genetics and genomics, including inferences of be based on the shared descent of genome to currently observable species, populations, or individuals. Realization of this shared descent of genome, jointly across the genome and among individuals provides likelihood estimates. The variation in latent variable realizations conditional on observed data provides estimates of likelihood uncertainty. This uncertainty is informative in interpretation of the likelihood surface, and to the inferences that can be drawn.