CMStatistics 2017: Start Registration
View Submission - CMStatistics
Title: Compound Cox processes applied to extreme meteorological events Authors:  Nuria Ruiz-Fuentes - University of Jaen (Spain) [presenting]
Paula Bouzas - University of Granada (Spain)
Carmen Montes-Gijon - University of Granada (Spain)
Abstract: Given the enormous socioeconomic and environmental impact of extreme meteorological events, studies that provide predictive data in this area have become increasingly prevalent in recent decades. A series of this type of events, such as extreme values of temperatures (maximum and/or minimum), high levels of ultraviolet radiation or precipitation in a climate zone, may be modeled as a point process or as the resulting counting process; the compound Cox process is proposed for this purpose. Examining available data from different climate zones from 1991 to 2011, their intensity processes were estimated by means of an ad hoc functional principal components model. Having obtained the estimated models, the counting processes can be forecast (i.e. their mean, number of points and the probability of having a new event in a chosen future interval of time within a year). In addition, using a goodness-of-fit test, a new sample path collected in any climate zone can be tested to ascertain if it follows the same stochastic model; thus, the hypothesis test helps to discriminate between climate zones.