This paper introduces a multivariate circular–linear (or poly-cylindrical) distribution obtained by combining the projected and the skew-normal. We show the flexibility of our proposal, its closure under marginalization, and how to quantify multivariate dependence. Due to a non-identifiability issue that our proposal inherits from the projected normal, a computational problem arises. We overcome it in a Bayesian framework, adding suitable latent variables and showing that posterior samples can be obtained with a post-processing of the estimation algorithm output. Under specific prior choices, this approach enables us to implement a Markov chain Monte Carlo algorithm relying only on Gibbs steps, where the updates of the parameters are done as if we were working with a multivariate normal likelihood. The proposed approach can also be used with the projected normal. As a proof of concept, on simulated examples we show the ability of our algorithm in recovering the parameter values and to solve the identification problem. Then the proposal is used in a real data example, where the turning-angles (circular variables) and the logarithm of the step-lengths (linear variables) of four zebras are modeled jointly.

The joint projected normal and skew-normal: A distribution for poly-cylindrical data / Mastrantonio, G.. - In: JOURNAL OF MULTIVARIATE ANALYSIS. - ISSN 0047-259X. - 165:(2018), pp. 14-26. [10.1016/j.jmva.2017.11.006]

The joint projected normal and skew-normal: A distribution for poly-cylindrical data

Mastrantonio G.
2018

Abstract

This paper introduces a multivariate circular–linear (or poly-cylindrical) distribution obtained by combining the projected and the skew-normal. We show the flexibility of our proposal, its closure under marginalization, and how to quantify multivariate dependence. Due to a non-identifiability issue that our proposal inherits from the projected normal, a computational problem arises. We overcome it in a Bayesian framework, adding suitable latent variables and showing that posterior samples can be obtained with a post-processing of the estimation algorithm output. Under specific prior choices, this approach enables us to implement a Markov chain Monte Carlo algorithm relying only on Gibbs steps, where the updates of the parameters are done as if we were working with a multivariate normal likelihood. The proposed approach can also be used with the projected normal. As a proof of concept, on simulated examples we show the ability of our algorithm in recovering the parameter values and to solve the identification problem. Then the proposal is used in a real data example, where the turning-angles (circular variables) and the logarithm of the step-lengths (linear variables) of four zebras are modeled jointly.
File in questo prodotto:
File Dimensione Formato  
2018 - The joint projected normal and skew-normal.pdf

non disponibili

Tipologia: 2a Post-print versione editoriale / Version of Record
Licenza: Non Pubblico - Accesso privato/ristretto
Dimensione 1.11 MB
Formato Adobe PDF
1.11 MB Adobe PDF   Visualizza/Apri   Richiedi una copia
JMVA_2017_77-EiC.pdf

accesso aperto

Tipologia: 1. Preprint / submitted version [pre- review]
Licenza: PUBBLICO - Tutti i diritti riservati
Dimensione 731.86 kB
Formato Adobe PDF
731.86 kB Adobe PDF Visualizza/Apri
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11583/2777034