Desempenho das simulações AR e MA na plataforma Livre R

• Autor(a) principal: Santos, Nilson Barros
• Data de Publicação: 2015
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da UFS
• Texto Completo: http://ri.ufs.br/jspui/handle/riufs/8339
• topic: Estatística; Ensino de estatística; Algorítmo; Monte Carlo; Séries temporais; Algorithm; Monte Carlo; Time series; CIENCIAS EXATAS E DA TERRA::PROBABILIDADE E ESTATISTICA::ESTATISTICA::ANALISE DE DADOS

This study aims to evaluate the performance of Autoregressive (AR) model simulations and Moving Average (MA) in the Free Platform R. Therefore, we use a set of procedures (functions and commands) in Free Software R, generating simulated series for different scenarios. Were simulated by algorithm, time series models: AR (1), AR (2), MA (1) and MA (2) varying the parameter values of each proposed model, the size of the series (were simulated series with n = 10, n = 20, n = 100, n = 500 and n = 1, 000 observations) as well, the amount of generated series for each value of the parameters is simulated between 90-9000 series in each run of the algorithm. The time series simulated by the arima.sim were analyzed automatically using by the function auto.arima and through observation the frequency of the generated time series, in line with the initially proposed model, conclude that the simulated series do not always correspond to the model originally proposed. Especially for a series of simulations with n small, i.e. with few observations. To better understand the work still approach the key concepts of time series modeling proposed by Box and Jenkins for stochastic processes - AR (p) and MA (q) and, finally the main concepts related to statistical simulation (Simulation Method Monte Carlo).