A0222
Title: A non-parametric method for high dimensional change point analysis
Authors: Lupeng Zhang - Durham University (United Kingdom) [presenting]
Reza Drikvandi - Durham University (United Kingdom)
Abstract: Change point analysis aims to detect significant changes in the distribution of a data sequence. It holds critical importance across modern statistical applications such as economics, finance, quality control, genetics and medical research. While change point detection for low dimensional data is extensively studied in the literature, change point detection is very challenging in high dimensional data where the number of variables is much larger than the number of observations. High-dimensional change point analysis has become a vital focus of recent research. We discuss the main challenges and difficulties with high dimensional change points and introduce a nonparametric approach to tackle some of those challenges. The proposed method is based on some dissimilarity distances and CUSUM statistics to detect significant change points in high dimensional data. Our method can detect changes in both mean and variance of high dimensional observations, as well as other distributional changes. We present simulation results and a real data application on the S\&P 500 data.
