Target Analysis Report — Difference Decomposition Analysis
Overview
The objective of this function is to analyze the difference between two scenarios, such as two products. It takes as inputs a Bayesian network with causal nodes and two sets of evidence for those causal nodes. It returns a decomposition of the overall difference between the two scenarios with respect to each causal node in the network, following the sequence of actions in topological order.
Causal Graph
A Bayesian network validated as causal by the analyst:
In this example, we compare two products, labeled 10-FA and 12-FM. Product 10-FA has an overall rating (“global rate”) of 7.758, while 12-FM has an overall rating of 6.9796. Setting [Factor_3] for 12-FM to the level of 10-FA increases the rating to 7.0764, and setting [Factor_4] for 12-FM to the level of 10-FA decreases the rating to 6.9832. As a result, the cumulative difference is 0.0036.
