Structural Breaks and Time Series Data: Comparison and Real Data Analysis

• Activity Number: 440 - SLDS CSpeed 8
• Type: Contributed
• Date/Time: Thursday, August 12, 2021 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistical Learning and Data Science
• Author(s): Arick Grootveld*
• Companies: Western Washington University
• Keywords: Structural Breaks; Change Point Analysis; Likelihood Ratio; Schwarz Information Criterion; Time Series

Abstract

Testing on structural change problems on dependent observations has been an important issue in statistics. Different methods have been developed to detect and estimate change(s) in time series data. In this presentation, we are going to use one popular method, Likelihood Ratio (LR) method to detect change(s)in structure of time series data. In order to use the LR method, we need to convert the dependent data problem into an independent data problem. We will discuss several approaches to convert a time series data into independent data. Using Monte Carlo simulations, we will compare LR method and Schwarz Information Criterion (SIC) method, which is a popular model selection criterion. Further, different error distributions and different sample sizes will be taken into consideration to make comparisons. Finally, we will analyze real data to estimate change(s) and their location(s) to illustrate the efficiency of the procedures.