(Universidade Federal do Espírito Santo, 2013-08-27) Paula Pinto, Wanderson de; Reisen, Valdério Anselmo; Sena Junior, Manuel; Albuquerque,Taciana Toledo de Almeida
Data of air pollution have generally missing observations. This research presents a study of methods to estimate the autocorrelation function in the presence of missing data, based on the work of Yajima and Nishino (1999). There is also some techniques for imputation of missing data based on the use of the EM algorithm proposed by Dempster (1977), and the ARIMA time series models of Box and Jenkins. Testing simulations with frame proportions of missing data were performed to compare the mean square errors of the proposed estimators. The empirical study showed that the proposed estimation method has good performance in terms of mean squared error measures. As an illustration of the proposed methodology, two time series of concentrations of Inhalable Particulate Matter (PM10) issued in the Region of Vitoria, ES, Brazil, are analyzed.