A less arbitrary method for inferring cause and effect: Generalization of a medical model

A method is introduced that was developed for medical research in order to distinguish between random changes and changes with reproducible causes in the natural state of an empirical system. The method differs from statistical inference in that probability is associated with relative frequency only when characterizing the natural state of a system. More generally, it is used to distinguish signal from noise. For the latter purpose, probability is scaled for the actual boundary conditions imposed by a system, and a nonlinear spectrum-like function is used to relate low probability to signal (equivalently, high probability to noise).<>