Banca de DEFESA: ALEXANDRE HENRIQUE QUADROS GRAMOSA

Uma banca de DEFESA de DOUTORADO foi cadastrada pelo programa.
STUDENT : ALEXANDRE HENRIQUE QUADROS GRAMOSA
DATE: 04/08/2023
TIME: 14:00
LOCAL: meet.google.com/oxg-jjqv-nbi
TITLE:

Bayesian models for analysis of zero-inflated extreme data

KEY WORDS:

Extreme Events, Zero Inflation, Bayesian Analysis, Precipitation, MCMC, Seasonality.

PAGES: 73
BIG AREA: Ciências Agrárias
AREA: Agronomia
SUMMARY:

It is of fundamental importance to have knowledge of the limiting result for modeling block maxima of size n, known as the Generalized Extreme Value (GEV) distribution, which is used in modeling extreme events. However, in these extreme data, an excessive number of zeros can occur, which hinders the analysis and estimation using the GEV distribution. The Zero-Inflated Generalized Extreme Value (ZIGEV) distribution was recently created to solve this issue, with the aid of its inflator parameter ω. One of the objectives of this work is to apply this distribution to daily precipitation data, transformed into blocks of monthly maxima. In these data, there may be months without precipitation, which are computed as zero. Time series from the mesoregions of the state of Pernambuco, in the northeastern region of Brazil, were analyzed. Some of them had a predominance of non-rainy months. However, the main objective of this work is the creation of the Seasonal Zero-Inflated Generalized Extreme Value (SZIGEV) distribution, a model to analyze the seasonality of extreme data inflated with zeros. In this case, precipitation data from the cities of Recife and Petrolina in the state of Pernambuco, and from the cities of São João do Piauí and Teresina in the state of Piauí, were used. In both analyses, inferences were made under the Bayesian paradigm, with parameter estimation performed through numerical approximations of the posterior distribution using the Markov Chain Monte Carlo (MCMC) method. The results of these applications, in line with the work of Gramosa et al. (2019), reinforced that the analyses and estimations made by the ZIGEV distribution, compared to the GEV distribution, were more accurate and had a better quality of fit, highlighting the importance of using ZIGEV to model extreme data, especially when they are inflated with zeros. However, when comparing the application of ZIGEV with the SZIGEV distribution, it was noticed that due to the seasonal behavior exhibited by the data under study, the results obtained by the seasonal model (SZIGEV) were better and more precise, emphasizing the relevance of this distribution for modeling extreme data inflated with zeros and exhibiting seasonality.

COMMITTEE MEMBERS:
Interno - ANTONIO SAMUEL ALVES DA SILVA
Externo à Instituição - FERNANDO FERRAZ DO NASCIMENTO - UFPI
Externo à Instituição - FIDEL ERNESTO CASTRO MORAES - UFRJ
Presidente - FRANK SINATRA GOMES DA SILVA
Interno - JOSIMAR MENDES DE VASCONCELOS