Causality inference of nominal variables: A statistical simulation method

Abstract

In present study I proposed a statistical simulation method for causality inference of nominal variables (i.e., categorical variables). A new correlation measure for nominal variables, association coefficient, is firstly proposed also. A statistical simulation method was developed to generate artificial data of nominal variables with known causality. The law was then drawn from the simulation analysis of the artificial data. For a set of data of two nominal variables, the randomization method was first used to test the statistical significance of the nominal correlation measure, and then the statistical simulation was used to determine the causality and its statistic significance of two nominal variables. Full Matlab codes of the method were presented.

Keywords

• causality
• inference
• correlation
• nominal variables
• contingency
• association
• randomization
• statistical simulation
• independent variable
• machine learning
• artitificial intelligence