Library
What is the difference between the effects (total & direct) coming from BayesiaLab and Structural Equation modeling (SEM). In SEM we get the direct effect of those variables that have direct causal linkages to the target node; but in BayesiaLab we get the same for all nodes (variables).
Quote 0 0
Library
In SEM, the effects are computed using regression. BayesiaLab allows computing Direct Effects and Total Effects by using probabilistic inference with Soft Evidence on the current network. Inference is carried out for all the variables that are “observable” (Node property – Cost), not necessarily the variables that are directly connected to the target variable.[*:357x431u]For Total Effects, there is no causal assumption, i.e. the information can flow from one variable to the Target if there is an open path. [/*:m:357x431u][*:357x431u]Direct Effects are computed by fixing the probability distributions of all the variables, except for Non-Confounders and unobservable Factors. The information flow could be incorrectly blocked by these fixing the Non-Confounders and Factors.[/*:m:357x431u]
Quote 0 0