Abstract:
|
A new approach to causal modeling, based on intrinsic probability concept, was recently proposed and published. Its main ideas have general character, but were expressed for qualitative variables, binary and alike. In this presentation two types of extensions are proposed: new arguments on behalf of the new types of causal modeling (in particular, demonstration of failure of traditional approaches based on examples with "ideal models"), and possible simple way to make realization of the same ideas of causal modeling for numerical, quantitative variables. New algorithms are illustrated on media planning and demographic data.
|