Decisions on the causal structure of vector autoregressions
Robert M. Kunst · 2001
This paper ful…lls a double purpose. Firstly, we aim at study-ing causality tests with respect to their power of classi…cation into the four main competing hypotheses of causal interaction among two variables. We allow for various empirically relevant modi…cations of the basic bivariate autoregressive frame. These experiments result in some tentative recommendations for the empirical researcher. Secondly, we present and advertise the use of the decision maps technique, which builds on Bayes testing and is in itself a powerful simulation tool that can be used for many other statistical decision problems, particularly with a time-series background. 1