Event Date
Speaker: Guo Yu (Assistant Professor, Statistics and Applied Probability, UC Santa Barbara)
Title: "Sparse and positive-definite estimation for large covariance matrices with repeated measurements"
Abstract: Repeated measurements arise in many areas such as epidemiology, medicine, psychology and neuroscience, where random variables are measured multiple times across different subjects. In such settings, dependence structures among random variables that are between subjects and within a subject may be different. Ignoring this fact may lead to misleading and questionable analytic results, which is practically misguided. In this paper, we study the problem of simultaneous sparse and positive-definite estimation for the between-subject and within-subject covariance matrices. We establish the convergence rates for our proposed between-subject and within-subject covariance matrix estimators under some regularity conditions. In general, the convergence rate for the within-subject covariance matrix estimator depends on the total number of the observations, while the convergence rate for our between-subject covariance matrix estimator is affected by the number of subjects and insensitive to the imbalance of the data. The finite-sample performance of the proposed estimators are illustrated numerically in comprehensive simulations and real data application.
Faculty webpage (links to UC Santa Barbara): https://www.pstat.ucsb.edu/people/guo-yu
Seminar Date/Time: Thursday April 20, 2023, at 4:10pm
Location: MSB 1147 (Colloquium Room)
Refreshments: 3:30pm (MSB 1147 courtyard)