Further Thoughts on Testing for Causality

Abstract

The ontological basis for causality testing must be some empirically interpretable system leading one to specify a logically valid a priori law that can be shown to exist. Such paradigms, as conventional linear and non-linear models, be they structural or time series, tend to embody contradictory assumptions that unfortunately make causal interpretations problematic. The paper demonstrates how these difficulties can be avoided using a stochastic-coefficient approach. The final part of this article is devoted to a discussion of probabilistic logic as a valid tool for scientific analysis and interpretation of causal relationships.