Title: Testing independence under local stationarity via the local empirical characteristic function
Authors: Guy-Niklas Brunotte - Otto-Friedrich-Universität Bamberg (Germany) [presenting]
Abstract: Locally stationary processes are non-stationary processes which can be locally approximated by stationary processes (so-called companion processes) allowing us to handle them appropriately by using tools for stationary processes. Kac's theorem shows that independence between random variables can be expressed essentially by their characteristic functions. This is the basic idea behind the construction of a consistent level-alpha test for independence via a local version of the empirical characteristic functions. A global test for pairwise independence of companion processes at the same points in time is introduced which evaluate the L2-distance between the joint local empirical characteristic function of two locally stationary processes and the product of the local empirical characteristic functions of these processes.