Introducing non-stationarity to wrapped gaussian spatial responses with an application to wind direction
- Autores: Nadja Klein, Thomas Kneib, Isa Marques
- Localización: Proceedings of the 35th International Workshop on Statistical Modelling : July 20-24, 2020 Bilbao, Basque Country, Spain / Itziar Irigoien Garbizu (ed. lit.), Dae-Jin Lee (ed. lit.), Joaquín Martínez Minaya (ed. lit.), María Xosé Rodríguez Álvarez (ed. lit.), 2020, ISBN 978-84-1319-267-3, págs. 159-164
- Idioma: inglés
- Enlaces
-
Resumen
Circular data, i.e., data consisting of observations on the unit circle, can be found across many areas of science, for instance meteorology (wind directions), biology (animal movement directions), or medicine. The special nature of such data means that conventional methods for non-periodic data are no longer valid. As a consequence the analysis of such data is more challenging and the literature scarcer. In this paper, we introduce a spatial model for circular data that allows for non-stationarity in the mean and covariance structure of random elds. For this, we use the computationally ecient stochastic partial di erential equation approach. Moreover, we develop tunable hyper-priors, inspired by the penalized complexity prior framework, that shrink the model towards a base model with stationary covariance function. The performance of the proposed model is analyzed in detail in a simulation study, with a strong focus on the properties of hyper-priors considered. Finally, we evaluate the ability of our approach to estimate wind-directions during a wind storm in Germany.
