ScholarWorks@성균관대학교

초록

This paper introduces the sparse additive matrix autoregressive (SAdd-MAR) model for analyzing high-dimensional matrix-valued time series data. The model extends MAR model by incorporating $L_1$ regularization for sparsity. It also incorporates cross-sectional dependencies in the matrix-valued time series. We also adapt estimation method of Zhang (2024) by incorporating sparse estimation and thresholding to enhance model accuracy and reduce computational cost. Extensive Monte Carlo simulations show that our proposed method performs well. We also apply our model to economic indicators from OECD countries, demonstrating its superior forecasting power compared to other models.

키워드