The effect of temporal aggregation on the estimation accuracy of ARMA models. Issue 10 (26th November 2018)

• Record Type: Journal Article
• Authors: Teles, Paulo; Sousa, Paulo S. A.
• Is Part Of: Communications in statistics. Volume 47:Issue 10(2018)
• Page Start: 2865
• Page End: 2885
• Publication Date: 2018-11-26
• Subjects: Aggregate time series -- ARMA models -- Estimation accuracy loss -- Mean absolute error -- MA models -- Mean absolute percentage error -- Parameter values -- Temporal aggregation; 37M10 -- 62M10; Mathematical statistics -- Periodicals; Mathematical statistics -- Data processing -- Periodicals; Digital computer simulation -- Periodicals; 519.5
• DOI: 10.1080/03610918.2017.1361978
• Languages: English
• ISSNs: 0361-0918

Abstract

ABSTRACT: Autoregressive Moving Average (ARMA) time series model fitting is a procedure often based on aggregate data, where parameter estimation plays a key role. Therefore, we analyze the effect of temporal aggregation on the accuracy of parameter estimation of mixed ARMA and MA models. We derive the expressions required to compute the parameter values of the aggregate models as functions of the basic model parameters in order to compare their estimation accuracy. To this end, a simulation experiment shows that aggregation causes a severe accuracy loss that increases with the order of aggregation, leading to poor accuracy.