Abstract:
|
A multivariate time series model, of the structural vector auto-regressive moving average type, can be represented by a graphical model where each node is a variable at a time index (t, t-1,... etc). In this context, edges would represent conditional dependence between the connected time series. While a saturated models includes edges between each variable at time t and all the other variables, a parsimonious (sparse) models would have some of the edges deleted. In the research we illustrate different approaches to test for the significance of the presence of the edges and compare them through simulation studies.
|