Avaliação do desempenho e aplicação do gráfico de controle EWMA para o processo estocástico POMINAR(1)

Abstract

In this work we propose to evaluate the performance of the EWMA control chart, adapting it to the first-order mixed autoregressive model with Poisson innovations, called POMINAR(1), which is composed of two operators known as thinning binomial and thinning Poisson. The objective is to compare the results obtained with those of the Shewhart control chart, and to show that the EWMA control chart is more efficient to detect deviations in the process mean below 1.5 standard deviation. For this, the EWMA control chart and its expectation are defined. Two theorems and two corollaries are proposed in order to define the variance according to the size of the rational subgroup. Moreover, the EWMA upper and lower control limits will be set, to analyze its performance. After that, an analysis will be made comparing the values found in a simulation using the EWMA control chart with the results obtained using the Shewhart control chart. Finally, two applications with real data using the EWMA control chart are presented. The first situation relates to the Criminal Data Section on the Forecasting Principles website. The chosen variable represents the number of monthly robberies reported in political district nº 12, in Pittsburgh (USA), from Jan/1990 to Dec/2001, with a total of 144 observations. The second situation is the count of different IP addresses recorded in a period of two minutes in duration on the server of the Department of Statistics of the University of Wüsrzburg, on November 29, 2005, between 10 a.m. and 6 p.m. with 241 observations.