Estimating Nonstationary Time Series with Missing Data

Abstract

A new method for estimating ARIMA model parameters of nonstationary time series with missing observations is considered. Two approaches are proposed: (i) a nonstationary time series is transformed into stationary one, a stationary time series missing data iterative estimation algorithm is applied and at its end the inverse transformation is performed to obtain the estimates of original time series missing observations, (ii) the inverse transformation and the calculation o original time series missing observations is performed in each loop of the iterative procedure and not only at the end. Use of the methods is illustrated by a case study.