This paper proposes modified Peter and Clark (PC) algorithm of graph-theoretic approach to study causality correlated data. The proposed algorithm is derived to determine the directions of the casual correlated complex variables. The PC algorithm treats VAR residuals as original variables while the proposed algorithm riz-PC uses modified R recursive residuals to find the correct causal direction among policy variables. This study evaluates the performance of these causal search algorithms in term of size and power properties. Our findings suggest that the newly proposed modified riz-PC algorithm can test causality better, as it successfully depicted the correct causal direction and was best at differentiating between true and spurious causality in routine Monte Carlo experiments.
展开▼
