Data Segmentation and High Dimensional Time Series Analysis

数据分割和高维时间序列分析

• 批准号: 2268657
• 金额: --
• 依托单位: University of Bristol
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2019
• 资助国家: 英国
• 起止时间: 2019 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2268657
• 关键词: Data; Segmentation; Dimensional; Time; Series

In my research, I will study time series data segmentation, particularly multiple change point analysis for multivariate time series in the hypothesis testing framework. I will first focus on developing a moving sum procedure for low-dimensional vector autoregression processes and deriving its statistical guarantees. This will then be extended to high-dimensional data, adopting a structured VAR model for computational feasibility. I will explore the applications of proposed methodologies to economic and financial data. There is potential to study their links to modern inference problems, such as sequential testing, and to network analysis. Later into the project, I may investigate other time series structures, such as volatility models, or other types of non-stationarity, such as cointegration. As a general theme, my aim is to develop a methodology for practitioners. Therefore, along with investigating into statistical guarantees of the proposed methodologies, I will also endeavour to develop fast and efficient algorithms and make their implementations available as software packages in the R language.

在我的研究中，我将研究时间序列数据分割，特别是假设检验框架中多元时间序列的多变点分析。我将首先专注于开发低维向量自回归过程的移动总和程序并推导出其统计保证。然后将其扩展到高维数据，采用结构化 VAR 模型来实现计算可行性。我将探讨所提出的方法在经济和金融数据中的应用。有潜力研究它们与现代推理问题的联系，例如顺序测试和网络分析。在项目后期，我可能会研究其他时间序列结构，例如波动率模型，或其他类型的非平稳性，例如协整。作为一个总体主题，我的目标是为从业者开发一种方法论。因此，在研究所提出的方法的统计保证的同时，我还将努力开发快速高效的算法，并以 R 语言软件包的形式提供它们的实现。