Temporal Indices

Context

In BayesiaLab, you can assign Temporal Indices to nodes. This allows you to define a temporal order between nodes, which learning algorithms take into account when a network is learned. For instance, a learning algorithm would not discover that node $A$ is a parent or ancestor of node $B$ if the Temporal Index of node $A$ is greater than the Temporal Index of node $B$.

Temporal Indices can take on any real values equal to or greater than 0. A Temporal Index of 0 refers to nodes associated with the oldest possible point on a timeline, while higher values indicate that nodes are associated with more recent points on that same timeline. However, no meaning is associated with the interval between individual Temporal Indices. For learning purposes, the Temporal Indices of 0 and 1 convey the same temporal ordering as the values of 0.2 and 0.3. By default, there are no Temporal Indices assigned to nodes.

Usage

There are several ways to edit Temporal Indices:

• Select Node Contextual Menu > Properties > Temporal Index and specify a value.
• Select Graph Panel Contextual Menu > Edit Temporal Indices to open the Temporal Indices Editor.
• Click on the Temporal Index indicator in the Status Bar to bring up the Temporal Indices Editor.

When setting a value from Node Contextual Menu > Properties > Temporal Index, the Propagate Index to the Classes button applies the same index to all nodes in the selected classes (or clears it if there is no index).

Example & Workflow Animation

For illustration purposes, we use the well-known Visit Asia example. Here, we assign a temporal ordering based on domain knowledge: demographic variables precede risk factors, which precede conditions, which come before symptoms. Furthermore, we use the Associate Color function in the Temporal Indices Editor and sort the table in ascending order.

Import/Export Temporal Indices

The Import and Export buttons in the Temporal Indices Editor allow you to manage Temporal Indices using Dictionaries.