{"id":"arxiv/2609.22444","url":"https://arxiv.org/abs/2609.22444","tags":["math.ST","math.PR","q-bio.PE","stat.ME"],"score":0,"title":"Asymptotics and finite sample bounds for prediction and smoothing in Wright-Fisher hidden Markov models","domain":"math.ST","source":"arxiv","authors":["Luigi M. Malgieri","Filippo Ascolani","Matteo Giordano","Matteo Ruggiero"],"reviews":0,"subject":"math.ST","summary":"We study signal prediction and smoothing in hidden Markov models whose latent signal is a multi-type Wright-Fisher diffusion and whose discrete-time observations are categorical. We consider a repeated-sampling regime with finitely many inspection times and large within-time samples. Under parent-independent mutation, the Bayesian predictive and smoothing laws admit exact representations as mixtures of Dirichlet distributions. We show that the one-step predictive and local smoothing laws converge in total variation to the corresponding Wright-Fisher transition and bridge laws determined by…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-09-18","reactions":{}}