Functional Data Analysis in Practice

Overview

About the Talk:

Functional Data Analysis (FDA) is a branch of statistics that focuses on the analysis of data in the form of curves and surfaces, or less regular functions such as trees and graphs. In this non-orthodox talk on FDA, we illustrate some added advantages of FDA over the conventional data analysis by three real data applications: (i) Absorbing nonstationarity into a stationary functional framework — Forecasting the daily electricity load curves. (ii) Estimating extreme quantiles for a random function based on a small sample — Testing for term structure pricing models. (iii) Clustering short time series based on common dynamic structures — Estimating food-chain-interaction between mink and muskrats.

Biography of Speaker:

Qiwei Yao is Professor of Statistics at London School of Economics and Political Science. His main research interest is on statistical inference for complex time series, including high-dimensional time series, dynamic networks, spatio-temporal processes, functional time series, and nonlinear time series. He was Saw Swee Hock Professor of Statistics (2022) at the National University of Singapore.