Comparison of automated procedures for ARMA model identification

Abstract

This article evaluates the performance of three automated procedures for ARMA model identification commonly available in current versions of SAS for Windows: MINIC, SCAN, and ESACF. Monte Carlo experiments with different model structures, parameter values, and sample sizes were used to compare the methods. On average, the procedures either correctly identified the simulated structures or selected parsimonious nearly equivalent mathematical representations in at least 60% of the trials conducted. For autoregressive models, MINIC achieved the best results. SCAN was superior to the other two procedures for mixed structures. For moving-average processes, ESACF obtained the most correct selections. For all three methods, model identification was less accurate for low dependency than for medium or high dependency processes. The effect of sample size was more pronounced for MIMIC than for SCAN and ESACF. SCAN and ESACF tended to select higherorder mixed structures in larger samples. These fmdings are confined to stationary nonseasonal time series.Time series analysis deals with repeated and equally spaced observations on a single unit. Classical statistical techniques are no longer appropriate here because data points cannot be assumed independent and uncorrelated. One of the most widely employed procedures for time series data is the autoregressive integrated moving-average (ARIMA) approach proposed by Box and Jenkins (1970). Since Glass, Willson, and Gottman (1975) introduced the ARIMA technique to the social and behavioral sciences, this methodology has been increasingly employed in different research fields (Delcor,

Trend-Residual Dual Modeling for Detection of Outliers in Low-Cost GPS Trajectories

“…Besides, for a long GPS trajectory, one may divide the trajectory into multiple segments. We suggest that one segment is at least 100 points to ensure correctly identifying time series model of residuals [32,33]. …”

Deterministic or Stochastic Trend

“…Employing the Box-Jenkins procedure, Glass and colleagues fitted the (0, 1, 1) model to the series. (For an elaborated strategy of model selection combining different techniques, see Stadnytska, Braun, and Werner (2008a, 2008b).) Therefore according to Glass et al (1975), a positive trend in the series is due to a stochastic drift…”