The Joint Projected and Skew Normal

TL;DR

This paper introduces a new multivariate circular-linear distribution tailored for animal movement data, incorporating heterogeneity and time dependence via a hidden Markov model, with Bayesian estimation and MCMC implementation.

Contribution

It presents a novel joint projected and skew normal distribution for modeling complex movement data, advancing the statistical tools available for animal movement analysis.

Findings

Model applied to Maremma Sheepdog data shows good predictive performance.

Compared favorably to models using traditional distributions in interpretability.

Demonstrates flexibility in capturing heterogeneity and temporal dependence.

Abstract

We introduce a new multivariate circular linear distribution suitable for modeling direction and speed in (multiple) animal movement data. To properly account for specific data features, such as heterogeneity and time dependence, a hidden Markov model is used. Parameters are estimated under a Bayesian framework and we provide computational details to implement the Markov chain Monte Carlo algorithm. The proposed model is applied to a dataset of six free-ranging Maremma Sheepdogs. Its predictive performance, as well as the interpretability of the results, are compared to those given by hidden Markov models built on all the combinations of von Mises (circular), wrapped Cauchy (circular), gamma (linear) and Weibull (linear) distributions