Online Program Home
My Program

Abstract Details

Activity Number: 350 - New Methods for Time Series and Longitudinal Data
Type: Contributed
Date/Time: Tuesday, July 30, 2019 : 10:30 AM to 12:20 PM
Sponsor: Section on Statistical Learning and Data Science
Abstract #304172
Title: Joint Estimation of Structured Multivariate VAR Modeling
Author(s): Peiliang Bai* and George Michailidis
Companies: University of Florida and University of Florida
Keywords: Multivariate VAR process; structured estimation

In this paper, we consider the estimation of structured multivariate VAR modeling. The big challenge is that as either the dimension of VAR process grows or the number of VAR variables grows, that the number of parameters grows extremely fast so that conventional lasso method would fail to estimate both the dependent effects among variables and the correlations within each VAR process are difficult to be measured. We proposed a new model for illustrating the dependencies within each variable and cross-influence among these VAR variables precisely, we proposed a procedure for estimating the correlation among variables and within each VAR process alternatively by using ADMM, we also proved the asymptotic properties like restricted strong bi-convexity and deviation bounds. We applied our model to analyze the climate defined variable database. Numerical studies and real data analysis demonstrate that both theory and algorithm work efficiently and outperform the conventional lasso estimation.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2019 program