Ruofan Yu, Rutgers University
Project Description/Abstract
Matrix time series are time series data observed in matrix form. Examples of matrix time series data could be import-export data between countries, or time-indexed images that are derived from videos. Matrix factor model assumes the movement is driven by some latent factors in matrix form, which reduces the dimensional of the data. In this talk, we introduce Dynamic matrix factor model that extends the Matrix factor model by bringing some dynamic structure into the latent factor process. We’ll present the estimation and properties of this dynamic matrix factor model, and the advantages of the model will be discussed and studied via making predictions on a real data example.
Co-authors
Yuefeng Han, Rutgers University
Rong Chen, Rutgers University
Han Xiao, Rutgers University