CAUSALITY TEST FOR NON-STATIONARY TIME SERIES

摘要

A proposal is made for a causality test for non-stationary time series. This is done by transforming the I.J Good's originally specified probabilistic causality theory from random events into a stochastic-process framework. With the adoption of a non-stationary framework, premise assumptions of the widely used and misused Granger Causality are violated. This paper suggests a model-free probability raising causality test which assumes no stationarity a priori. The proposed methodology is validated via simulations.