Child pages
  • 2D Mapping (7.0)

Contents

Context

Analysis | Visual | Overall | Mapping | 2D Mapping

History

Mapping was first introduced in version 5.1, and then updated in version 5.3 and 5.4. After the addition of a 3D version, this feature has been renamed 2D Mapping.

New Node Analyses

Normalized Entropy

Introduced in version 7.0, Normalized Entropy is available for defining the size and/or color of the nodes.

Evidence Analysis

We have added four metrics for measuring the changes between the posterior distributions of  given the set of observations  and the original set of evidence  (potentially empty):

  1. Bayes Factor
  2. Relative Bayes Factor
  3. Kullback-Leibler
  4. Relative Kullback-Leibler

 corresponds to the set of evidence that is defined upon selecting the Node Analysis.

 is defined by using the Monitors associated with the nodes.


Example

Let's use our familiar VisitAsia example to describe these metrics. We are assuming that  corresponds to the empty set of evidence.

Four each of these four metrics, we use two sets of evidence :

  1. the patient is a smoker,
  2. the patient is a smoker and has some difficulties to breath.

This metric is the sum, across each state of the variable, of the difference between the log-losses of the two posterior probabilities

 

 

 

 

This metric is the Bayes Factor divided by the number of states of the variable

 

This metric is the sum, across each state of the variable, of the symmetrical Kullback-Leibler divergence between the two posterior probabilities

 

This metric is the Kullback-Leibler divided by the number of states of the variable

These four metrics are all based on log-losses. However, as the KL metrics weight them with the probabilities, they are less sensitive than the Bayes Factors.

 


New Feature: Layout Transfer

Clicking the icon  allows transferring the current layout of the graph in the Mapping window to the Graph panel and thus change the layout of the Bayesian network.

 

New Feature: Adjacency Matrix

A matrix can now be generated to describe the current structure of the network displayed in the Mapping:

  • The rows represent the parents, 
  • The columns represent the children
  • The values in the cells depend on the selected Arc Analysis mode.