“Statistical Analysis of Tensor Time Series” by Dr. Dan Yang
Dr. Dan Yang
Department of Statistics & Biostatistics
Rutgers, The State University of New Jersey
Modern data collection capabilities have lead to massive quantity of time series. Large tensor (or multi-dimensional array) data are now routinely collected in a wide range of applications, and often such observations are taken over time, forming tensor time series. Although it is natural to turn the tensor observations into a long vector then use standard vector time series models, it is often the case that the fibers of a tensor represent different sets of information that are closely interplayed. We propose novel auto regressive and factor models for tensor time series that maintain and utilize the tensor structure to achieve greater dimensional reduction as well as more interpretable results. Estimation procedures and their theoretical properties are further investigated and demonstrated with simulated and real examples.