Hidden Markov models: Pitfalls and opportunities in ecology

Hidden Markov models (HMMs) and their extensions are attractive methods for analysing ecological data where noisy, multivariate measurements made of a hidden, process, this hidden process is represented by sequence discrete states. Yet, as these become more complex challenging to understand, it important consider what pitfalls have opportunities there future research address pitfalls. In paper, we review five lesser known one can encounter when using HMMs or solve problems: (a) violation the snapshot property in continuous-time HMMs; (b) biased inference from hierarchical applied temporally misaligned processes; (c) sensitive random effects partially pool across heterogeneous individuals; (d) computational burden approximate with continuous state spaces; (e) difficulty linking space environment. This ecologists statisticians familiar HMMs, but who may be less aware problems that arise specialised applications. We demonstrate how each pitfall arises, simulation example, discuss why consider. Along identifying problems, highlight potential offer ideas help alleviate Each solutions current problems. intend paper heighten awareness applying advanced methods, also hope highlighting opportunities, inspire weaken provide improved methods.