Outlier detection in structural time series models: The indicator saturation approach

Abstract

Structural change affects the estimation of economic signals, like the underlying growth rate or the seasonally adjusted series. An important issue, which has attracted a great deal of attention also in the seasonal adjustment literature, is its detection by an expert procedure. The general-to-specific approach to the detection of structural change, currently implemented in Autometrics via indicator saturation, has proven to be both practical and effective in the context of stationary dynamic regression models and unit-root autoregressions. By focusing on impulse-

Outliers detection in state-space model using indicator saturation approach

“…Hence, this gives a total of − ( − 1) 2 ⁄ runs of the selection procedure. This approach was employed by [14] on the BSM context. Inspired by their work, we integrated the IS approach in the local level model as in (5).…”

“…− Size of an outlier is given as where s is a positive integer and σ is the prediction error standard deviation (PESD) of the series. As a reference to [14], we set 7σ as benchmark value that determines the size of an outlier. However, the size of outlier varies from 3σ, 5σ, 9σ and 12σ.…”

“…This will ensure the low gauge value below 5% of the sample, T or only one irrelevant indicator variable is maintained in the final model. On the other hand, the concept of potency and gauge used in this study can be illustrated as a confusion matrix as in [14]. However, the following confusion matrix only summarizes the result of one Monte Carlo experiment shown in Table 1.…”

“…At present, research about the performance of indicator saturation especially from structural time series viewpoint is done by [14] using basic structural model (BSM) only. The work presented in this paper is motivated by the fact that outliers can be identified and modelled via the indicator saturation approach.…”

“…Prior works in [21] and [22] do not employ IS in Gaussian state-space model. Meanwhile, [14] applied impulse and step indicators to detect outliers and level shift in the basic structural model. Our study attempts to contribute to the literature by examining the performance of IIS in the context of the local level model by assessing the performance of IIS in LLM through Monte Carlo simulations replicated at = 1000 times.…”