On Gibbs Sampling for Endpoint-Conditioned Neighbor-Dependent Sequence Evolution Models
Yongkang Li, Joseph Mathews, Scott C. Schmidler · Journal of Computational and Graphical Statistics · 2025
Hobolth (2008) describes an algorithm for sampling paths in models of DNA sequence evolution with neighbor dependence. We show that this algorithm fails to sample the desired target distribution, and provide a simple Metropolis acceptance step correction. We also compare to a simpler Metropolis proposal, which appears to be more efficient and which we recommend instead. We demonstrate the impact of the correction on a parameter estimation problem studied in the original paper.