A Novel Structural AR Modeling Approach for a Continuous Time Linear Markov System

摘要: We often use a discrete time vector autoregressive (DVAR) model to analyse continuous time, multivariate, linear Markov systems through their series data sampled at steps. However, the DVAR has been considered not be structural representation and hence have bijective correspondence with system dynamics in general. In this paper, we characterize relationships of its corresponding AR (SVAR) (CVAR) models finite difference approximation differentials. Our analysis shows that bijectively corresponds dynamics. Further clarify SVAR CVAR are uniquely reproduced from under highly generic condition. Based on these results, propose novel Continuous Structural Vector AutoRegressive (CSVAR) modeling approach for derive empirically derived observed series. demonstrate superior performance some numerical experiments both artificial real world data.