Integrated Continuous-time Hidden Markov Models

TL;DR

This paper introduces integrated continuous-time hidden Markov models that account for entire interval dependencies of observations, providing a more efficient inference method applicable to movement ecology and related fields.

Contribution

It proposes a novel class of models that generalize existing switching diffusion models, enabling efficient likelihood evaluation via a conventional HMM framework.

Findings

Efficient likelihood evaluation using the Forward Algorithm.

Better scalability with data size compared to existing methods.

Successful application to animal movement data and simulations.

Abstract

Motivated by applications in movement ecology, in this paper I propose a new class of integrated continuous-time hidden Markov models in which each observation depends on the underlying state of the process over the whole interval since the previous observation, not only on its current state. This class gives a new representation of a range of existing models, including some widely applied switching diffusion models. I show that under appropriate conditioning, a model in this class can be regarded as a conventional hidden Markov model, enabling use of the Forward Algorithm for efficient evaluation of its likelihood without sampling of its state sequence. This leads to an algorithm for inference which is more efficient, and scales better with the amount of data, than existing methods. This is demonstrated and quantified in some applications to animal movement data and some related simulation experiments.