学术报告：夏强 华南农业大学

主讲人：夏强 华南农业大学

题 目：Statistical Inference of Rank-Decomposition-Based Matrix Factor Model for High-Dimensional Time Seriest

时间：2025年10月18日 15:50-17:20

地点：VSport体育官网新校园 B513

报告人：夏强 华南农业大学

摘要：This study examines the statistical inference of the hybrid structure two-way factor model (Hs2wFM) for high-dimensional matrix-variate time series based on rank decomposition. For this novel model, a two-step approach is proposed to estimate the loading matrices through eigenanalysis of a non-negative definite matrix derived from the autocovariance matrices. To determine the number of factors, we introduce two innovative ratio-based methods that utilize maximum- and sum-type statistics to assess the serial uncorrelatedness of the time series projections onto specific spaces. Theoretical properties, including convergence rates of estimated loading matrices and consistency of the ratio-based methods, are established under regularity conditions for the proposed model. The findings reveal that the convergence rates of the estimated loading matrices using the two-step approach are marginally superior. The effectiveness of the proposed method was validated through comprehensive simulations and analysis of real data.