摘要
Substantial research has delved into extracting the factor structure of the U.S. stock market, and many papers utilize the intraday returns since their availability. Employing the recently developed Tensor Principal Component Analysis method (TPCA introduced by Babii et al. in 2022), this paper tries to derive the underlying factors of intraday five-minute returns from the permanent constituents of the S&P 100. However, the data seem to suffer a small sample, strong intraday heteroskedasticity, and large error. In simulation, such problems can lead to insensible TPCA results. I propose a weighted version of TPCA that can potentially mitigate the issue and preserve the asymptotic properties and test implications of the TPCA.
类型
