This paper discusses decision rules, simulation, etc. describing my R package called "ganeralCorr". You can access it here, on SSRN.
It features my Keynote address given by me at the Jammu, India, few months ago. "52dn annual conference of the Indian Econometric Society." The paper has an Appendix with R code which shows how few lines can be used by data scientists to decide which variable is most likely to be the cause. For example, if we have data matrix mtx with columns X1, X2, X50 named so that X1 is the effect or response which we are most interested in. All other X's are possible causes.
Then R function "causeSummary(mtx)" will consider each variable paired with X1 and tell if Xj causes X1 or vice versa. The software is very sophisticated using non-parametric kernel methods. It can allow a set of variables to be "control variables" in all binary comparisons. Its implementation is very easy and kind of fun according to my students. It has tremendous potential in solving all kinds of scientific problems.
About the document
30 Pages Posted: 11 Jul 2018. Author: Hrishikesh D. Vinod, Fordham University - Department of Economics.
Statistically dependent X and Y have conditional density f(Y|X) asymmetrically different from unconditional f(Y). Theorem 1 com- pares two flipped kernel regressions to derive necessary conditions for causal path X --> Y and exogeneity meaning that X is self-driven. Hausman-Wu's indirect exogeneity test diagnoses a disease (endogeneity) by showing that instrumental variables (IV) estimator remedy 'works.' Instead, a unanimity index-based test aggregates evidence from four orders of stochastic dominance and new asymmetric partial correlation coefficients to determine the direction and strength of causal and exogenous variables. A simulation supports our decision rules. An illustration identifies exogenous variables which can help predict US economic recession.
Keywords: Kernel Regression, stochastic dominance, statistical independence, recession forecasting.