Library
Is it correct that reflective factors cannot be used for simulation?
Quote 0 0
Library
Observational inference in Bayesian networks allows setting evidence on manifest variables and simulating the impact on the latent variable or factor. However, a piece of evidence on a manifest variable will also be propagated to the other manifests within the same construct. Total Effect:We compute the effect of increasing the value of the manifest variable Trust by 1.totalEffect.png Marginal DistributionsmarginalDistributions.png Posterior DistributionsposteriorDistributions.png The total effect of Trust on [Factor_0] is +0.799. However, there is also an effect of +0.796 on the manifest variable Bold, +0.790 on Fulfilled, etc.In order to prevent these collateral effects, and to get the effect of one manifest only, it is possible to fix the probability distribution of the other manifests?Direct Effect:[*:2fxle7ex]We fix the distributions of all the other manifests by using the Contextual Menu associated with the Monitor.[/*:m:2fxle7ex][*:2fxle7ex]We compute the effect of increasing the value of the manifest variable Trust by 1.[/*:m:2fxle7ex]Monitor's Contextual MenucontextualMenu.png Marginal DistributionsmarginalDistributions1.png Posterior DistributionsposteriorDistributions1.png The direct effect of Trust is +0.243 on [Factor_0], and there is no effect on the other manifests.
Quote 0 0