Relationship with Target Node (9.0)

Context

This function, which was called in previous versions Correlation with Target Node, generates a comprehensive report with a lot of metrics related to the Target node. This report is also generated at the end of Data Clustering for analyzing the created latent variable.

Usage

Select Menu > Analysis > Report > Target > Relationship with Target Node

New Feature: Max and Min Bayes Factors

The tables describing the analyzes of the nodes with each Target State have been simplified.

The Maximal Positive and Negative Variations are now replaced by Bayes Factors:

$\displaystyle BF(e, s) = \frac{P(e \mid s)}{P(e)}$

where $e$ is a state of the node under analysis, and $s$ is the Target State.

Example

Let’s use the network below and analyze the relationship of the nodes with the target state $s$, the fifth state of the Target Node, $\mathit{Price\ (USD)} \gt 1140$.

We have highlighted, for the first and last nodes, $\mathit{sqft\_living}$ and $\mathit{Age}$, the metrics that are both in the table and in the monitors below:

• the Posterior Mean Value (light green),
• the posterior probability of the state $e_M$ that has the Maximum Bayes Factor $BF(e_M, s)$ (green)
• the posterior probability of the state $e_m$ that has the Minimum Bayes Factor $BF(e_m, s)$ (red)

New Feature: Color-Coding of the States

The states are now color-coded in the report tables to allow quick identification of their ranks, from red, for the first state, to green, for the last one.

Example