Abordagem Bayesiana para distribuição das r-maiores estatísticas de ordem (GEVr) com estrutura de modelos dinâmicos

Abstract

In series a collection of observations made sequentially over time. This type of change is Common for data applied in the theory of extreme values (EVT). In environmental data, for example, in rain, wind and temperature, Their levels may be correlated with seasonality, in addition to showing a tendency to increase over the Due to climate change on the planet. Generally, this type of event has been worked on Using standard parametric distributions such as Normal or Gamma, look at Camargo et al. (1994). However, environmental data, in most cases Cases have a heavy tail, unlike these distributions. In some situations (EVT) Analyzing only the generalized extreme value distribution (GEV) of a set of data can provide few Observations, in these cases it is more interesting to use the distribution of r-largest order statistics (GEVr). This work consists of the development of an algorithm in Software R for posterior distributions for GEVr based on the Bayesian estimation using Markov chains (MCMC) and the use of the Metropolis-Hastings algorithm technique. A Dynamic Linear Model (DLM), which is a general class of time series models, has also been introduced to model the GEVr parameters over time. The proposed model was applied in the time series of the temperature in ºC Teresina-PI and return BOVESPA , in order to follow the seasonality of the temperature in the capital of Piauí and level return BOVESPA, also incorporated was a Linear Dynamic Seasonal Model (DLMS), which is a class of time series models, for the GEVr parameter model over time. The proposed model was applied in the time series of temperature of ºC Teresina-PI, Curitiba-PR and Brasília-DF.