This paper proposes a method to construct well-calibrated frequentist prediction regions, with particular regard to the highest prediction density regions, which may be useful for multivariate spatial prediction. We consider, in particular, Gaussian random fields, and using a calibrating procedure we effectively improve the estimative prediction regions, because the coverage probability turns out to be closer to the target nominal value. Whenever a closed-form expression for the well-calibrated prediction region is not available, we may specify a simple bootstrap-based estimator. Particular attention is dedicated to the associated, improved predictive distribution function, which can be usefully considered for identifying spatial locations with extreme or unusual observations. A simulation study is proposed in order to compare empirically the calibrated predictive regions with the estimative ones. The proposed method is then applied to the global model assessment of a deterministic model for the prediction of PM10 levels using data from a network of air quality monitoring stations.

Calibrated prediction regions for Gaussian random fields

Vidoni, Paolo
2018

Abstract

This paper proposes a method to construct well-calibrated frequentist prediction regions, with particular regard to the highest prediction density regions, which may be useful for multivariate spatial prediction. We consider, in particular, Gaussian random fields, and using a calibrating procedure we effectively improve the estimative prediction regions, because the coverage probability turns out to be closer to the target nominal value. Whenever a closed-form expression for the well-calibrated prediction region is not available, we may specify a simple bootstrap-based estimator. Particular attention is dedicated to the associated, improved predictive distribution function, which can be usefully considered for identifying spatial locations with extreme or unusual observations. A simulation study is proposed in order to compare empirically the calibrated predictive regions with the estimative ones. The proposed method is then applied to the global model assessment of a deterministic model for the prediction of PM10 levels using data from a network of air quality monitoring stations.
File in questo prodotto:
File Dimensione Formato  
pred_grf_Env_rev2.pdf

embargo fino al 01/04/2019

Descrizione: Articolo
Tipologia: Documento in Post-print
Licenza: Creative commons
Dimensione 782.21 kB
Formato Adobe PDF
782.21 kB Adobe PDF Visualizza/Apri
Lagazio_et_al-2018-Environmetrics.pdf

non disponibili

Tipologia: Versione Editoriale (PDF)
Licenza: Non pubblico
Dimensione 1.21 MB
Formato Adobe PDF
1.21 MB Adobe PDF   Visualizza/Apri   Richiedi una copia

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: http://hdl.handle.net/11390/1143919
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? 0
social impact