A SIMULATION STUDY ON ESTIMATING
TIME SERIES ARIMA MODELS BY SPLINES
Abstract. Prediction and interpolation may be considered as two major purposes of time series analysis. Selecting a proper model in time scope as a member of ARIMA models is an important task and is required many steps to obtain the proper model. Some nonparametric regression methods such as splines have many applications in various fields. In this article, spline methods are applied to estimate time series models in a simulation study. In the simulation study, some data sets are generated of various ARIMA models. Then, the basic ARIMA model that is considered for generating data is fitted to each of the data sets as the proper model and the fitness of the models are investigated. Besides, smoothing spline method is applied for obtaining the proper pattern of the same data sets. Furthermore, fitness of these methods is compared by Sum of Square Errors (SSE) criterion to determine the more appropriate method and determining performance of smoothing spline.
AMS Subject classification: 62M10


Download full article from here (pdf format).

DOI: 10.12732/ijam.v27i5.1

Volume: 27
Issue: 5
Year: 2014