Across Boundary Causality and Modeling Study

Author:Yin Fei

Supervisor:Ma Lei


Degree Year:2018





Causality across population boundaries is a study on whether known causation relationship could be established for another pupulation.This is a problem of extrapolation,whose important issue is the quantitative description,and whose difficulty is that the investigator is unable to control all the factors of the background.It is the problem of “unknown background factors”,which resulted in that effect produced by causal factors could not be defined across populations.To deal with this problem,Nancy Cartwright and Jeremy Hardie provided evidence-based linear model,which could solve the problem of selected causal factors.Daniel Steel constructed a single path Mechanism’s model,which tended to the effect step by step,but this model was not suited for the negative effective factors.Judea Pearl and Elias Bareinboim investigated the selection diagram which was based on the causal transportability,but causal conditions could not be transmitted by the causal transportability.These models both have their own advantages,but also have defects.So we should adopt a new visual angle to develop an extrapolation model which enable us to conquer the new empirical problems and to solve the problem of extrapolation.In the coordinate rationality view,models could be compared with each other.And by synthesizing the relevant advantages,a new causal mediation model can be formed.The causal mediation model adopted mediated technology to invent the M-factor,which could explain the multiple paths of causality.The causal mediation model produced multiple paths of TE,DE and IE which could be calculated by contribution probability,and also be related with linear model.To some extent,this could solve the problem of “unknown background factors”.In the new causal mediation model,the M-factor played the most important role,which became the core factor in the coordinate rationality theory.The method of causal modeling could become a progress of science.