Elimination of third-series effect and defining partial measures of causality (Q2759337)

This paper deals with defining the partial causality between a pair of processes based on the elimination from both of the processes only of the one-way effect due to the third process. The proposed elimination method would preserve the inherent feedback structure in the two processes concerned. The construction of the partial causal measures between a pair of non-deterministic stationary processes on the base of the elimination from this pair of the one-way effect by a third process is presented. The author extends \textit{C. Sims}' characterization [Am. Econ. Review 62, 540-542 (1972)] of \textit{C. W. J. Granger}'s non-causality [J. Econ. Dynamics Control 2, 329-352 (1980)] to the case of a third-series presence and shows that the proposed partial concept avoids \textit{C. Hsiao}'s spurious causality [in: Time Series Analysis: Theory and Practice. I (ed. O. D. Anderson), 671-698 (1982)]. Then the author extends the partial causal measures to the class of possibly non-stationary reproducible processes and shows how to construct those measures in non-stationary case. The discussion on large-sample inference on the partial measures is presented.