Activity day:2023-11-03
Published At:2023-10-24
Views:603
2023-10-24 updated
【 11 月 3 日 CRETA Seminar】
時間: 2023 年 11 月 3 日 (週五) 下午 2:00~3:30
地點:國立臺灣大學管理學院二號館 103 教室
講者:殷壽鏞教授 (國立臺北大學經濟學系)
演講主題:Three-pass regression filter with weak factors
講題摘要:
In this paper, we delve into the behavior of the three-pass regression filter within the context of time series forecasting, specifically considering the inclusion of weak factors. The estimates of slope coefficients for numerous observed predictors exhibit distinct asymptotic properties contingent on the strength of these factors. We observe that when the factors are weak, the three-pass regression filter lacks consistency. Nevertheless, this issue can be addressed when there's a sufficient number of weak factors. Additionally, we explore the impact of the ratio of weak factor strength to the number of weak factors. Our theoretical findings are substantiated by Monte Carlo simulations. Subsequently, we apply this methodology to scrutinize the predictive contributions to the U.S. industrial production index, comparing the outcomes with those from a shrinkage approach and boosting regression.