Bayesian modelling of marked point processes with incomplete records: volcanic eruptions. Issue 1 (8th October 2019)

• Record Type: Journal Article
• Authors: Wang, Ting; Schofield, Matthew; Bebbington, Mark; Kiyosugi, Koji
• Is Part Of: Journal of the Royal Statistical Society. Volume 69:Issue 1(2020)
• Page Start: 109
• Page End: 130
• Publication Date: 2019-10-08
• Subjects: Bayesian residual analysis -- Forecast -- Global volcanic eruption record -- Hazard -- Missing data; Statistics -- Periodicals; 519.5
• DOI: 10.1111/rssc.12380
• Languages: English
• ISSNs: 0035-9254

Abstract

Summary: Modelling point processes with incomplete records is a challenging problem, especially when the degree of record completeness varies over time. For volcanic eruption records, we expect the degree of missingness to depend on both the time and the size of an eruption. We propose a time‐varying intensity function for a marked point process to model the non‐stationary variation of the observed point process caused by missing data. We apply this model to global and regional volcanic eruption records and use Bayesian inference to obtain hazard estimates and their uncertainties based on the observed incomplete records, to carry out residual analysis and to provide forecasts.