Conference Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 425 - Nonparametric Methods for Dependent Data
Type: Contributed
Date/Time: Wednesday, August 10, 2022 : 10:30 AM to 12:20 PM
Sponsor: Section on Nonparametric Statistics
Abstract #322458
Title: Sample Splitting Based Inference for Multi-Dimensional Parameter in Time Series
Author(s): Yi Zhang* and Xiaofeng Shao
Companies: University of Illinois at Urbana-Champaign and University of Illinois at Urbana-Champaign
Keywords: Self-normalization; time series

The traditional testing procedure for the null hypothesis about a multi-dimensional parameter in a stationary time series requires a consistent estimator of the long run covariance matrix. The commonly used estimator needs to choose a bandwidth parameter and is difficult to implement in practice. The self-normalization (SN) method avoids the bandwidth choice, is asymptotically distribution-free under the null but the size distortion is serious when the dimension is moderate and the temporal dependence is strong. In this talk, I will present a sample splitting based approach to testing and show the favorable size performance through simulations.

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

Back to the full JSM 2022 program