Um estudo comparativo para modelos de séries temporais de contagem

Abstract

In this work, two methods for time series of counts are evaluated, the Autoregressive Moving Average Generalized Model (GARMA) and the Autoregressive Moving Average Model Generalized Linear Model (GLARMA). The main objective is to analyze the quality of t of the above models, using algorithms implemented in the R language. Another objective is to compare these models to the Generalized Linear Model (GLM), which allows the t of non Gaussian observations, but does not take into account the time dependence of such data. A simulation study is conducted in order to verify the behavior of the estimates. Two applications to real series are performed, the number of companies that went bankrupt in the United States in the years 1985-2012, and monthly number of cases of polio in a hospital. The models describe satisfactorily the behavior of the series.