本系logo
東海大學統計學系--演講公告-【學術演講公告】111年3月29日(二)

【學術演講公告】111年3月29日(二)

  • 單位 : 統計學系
  • 分類 : 演講公告
  • 點閱 : 156
  • 日期 : 2022-03-18
時 間:111年3月29日(星期二下午14:10~15:00)
地 點:管理學院新大樓M242
主講人:冼芻蕘教授 (清華大學經濟學系 教授)
講 題:Unified HAR inference for nonstationary linear processes of possibly heavy-tailed noises

Abstract
 
Over the last twenty years there has been an interest in non-stationary time series with infinite-variance noises. Using the probability theories developed in Zhang and Chan (2021, Econometric Theory 37, 892-925), we devise a residual-based block bootstrap (RBB) procedure for inference on non-stationarity. Unlike the wild bootstrap (WB) procedure suggested in Cavaliere, Georgiev, and Taylor (2018, Econometric Theory 34, 302-348), this procedure is heteroscedasticity- and autocorrelation-robust (HAR), while it also allows for heavy-tailed noises. The limiting distributions of our test is first shown, under (i) various tail index; and (ii) conditional heteroscedasticity. We then propose an m-out-of-n RBB which can be applied to both finite-variance and infinite-variance cases. Intensive simulation studies compare the RBB with the WB, the conventional DF test, the Phillips-Perron (PP) test, and the ADF-GLS test. We close this paper with empirical examples time series with infinite-variance noises. In sum, this paper provides a general treatment of nonstationary linear processes with GARCH/IGARCH noises.

Keywords and phrases:
Heavy-tailed noises; nonstationary linear processes; heteroscedasticity and autocorrelation-robust (HAR); residual-based block bootstrap (RBB); tail index; wild bootstrap (WB)