Evidence

The term Evidence refers to information that is currently known about the variables in a Bayesian network or a Dynamic Bayesian network.

The Evidence window allows each variable to be set to a specific value, or left as missing (Unknown/Null).

It also allows Time Series/Sequence data to be set on Temporal variables in a Dynamic Bayesian network. Again, values in a Time Series/Sequence can be missing.

NOTE

Evidence can also be set (on certain node types) directly in network viewer by clicking a check box next to a discrete node state, or next to the mean on a continuous node, or clicking in a chart. To set evidence on multiple states within the same variable hold down the Ctrl key.
A particular state can be excluded by holding down the Alt key. An intervention can be set by holding down the Shift key.

NOTE

Another way to set evidence is from Data explorer, using data sources such as databases and spreadsheets.

Evidence

Opening

The following operations, launch the Evidence window:

Missing Data

Data in both Bayesian networks and Dynamic Bayesian networks, can be missing. The term missing, is equivalent to unobserved, unknown or null.

If a value is not displayed for a variable, this implies the current state is missing.

To set a discrete variable to missing, either select the empty drop down item, or use the Delete key.

To set a continuous variable to missing, either clear the existing text, or use the Delete key.

Soft evidence

Bayes Server allows soft evidence to be entered on discrete variables in both Bayesian networks and Dynamic Bayesian networks.

Soft evidence, specifies that the exact state of the discrete variable is unknown, however information exists about the chances of the variable being in particular states.

To enter soft evidence, click the drop down next to the discrete variable, and select Soft Evidence. This will launch the Soft evidence window.

Temporal Evidence

For temporal variables in a Dynamic Bayesian network, evidence can be set on the same variable at different times. i.e. Time Series or sequence data.

NOTE

Note that time is zero based in Bayes Server. (i.e. the first three times are t=0, t=1 and t=2).

To change the time period, enter the series count in the Time Count text box, and click Update Time Count. Note that this will truncate any existing evidence.

Temporal evidence

Interventions

For information on using interventions, please see Interventions.

Copy

Evidence can be copied to the clipboard by selecting cells in a grid, and clicking the Copy button.

NOTE

The keyboard shortcut Ctrl+C can also be used. To select all data in a grid, first select a cell, and then use the keyboard combination Ctrl+A.

Cut

Evidence can be removed and copied to the clipboard by selecting cells in a grid, and clicking the Cut button.

NOTE

The keyboard shortcut Ctrl+X can also be used. To select all data in a grid, first select a cell, and then use the keyboard combination Ctrl+A.

Paste

The Paste button, allows tab delimited data to be copied into the Evidence window.

NOTE

The keyboard shortcut Ctrl+V can also be used.

NOTE

For temporal data, different times should appear on separate lines.

If the Indexes check box is checked, Paste expects any discrete evidence to be specified using the zero based index of each state, rather than the state name.

Clear

The Clear drop down, removes evidence, temporal evidence or both.

NOTE

Evidence can also be removed, by selecting cells, and using the Delete key.

Indexes

When the Indexes checkbox is checked, discrete evidence values are displayed using the zero based index of the state, instead of the name.

NOTE

This setting also effects Cut/Copy and Paste operations.