The Status Bar at the bottom of each Graph Window contains a wide range of indicators. This screenshot shows all of them "lit up" simultaneously. In practical use, you may only see a few of them at any given time.
A Graph Window is always in one of two possible modes, the Modeling Mode or the Validation Mode.
You can switch between these modes by clicking the Mode Buttons in the lower lefthand corner of the Graph Window.
- To switch to Modeling Mode (keyboard shortcut: F4)
- To switch to Validation Mode (keyboard shortcut: F5)
Depending on the inference options, switching into the Validation Mode may take some time as node probabilities have to be computed.
Certain processes, e.g. learning, inference, automatic positioning of the nodes, analysis, etc, will "block" some functionalities of the active Graph Window. A signal at the the left of the progress bar indicates the current state of system availability:
- Green, when the system is ready.
- Red, when a task is in progress. The progress bar is a qualitative indicator of the time spent on the currently active task. A click on the red signal stops the current task.
- Amber, while a task is being stopped and the network is returning to its normal state. Although you may be tempted to click the button multiple times "to make it stop faster", the process has to run its course and repeated clicking won't have any effect.
Among the standard items in the Status Bar is the Selection Counter. It displays the number of nodes and arcs that are currently selected in the graph window:The node count is shown to the left of the colon; the arc count is on the right.
By default the Structural Coefficient, also denoted by the Greek letter alpha (α) is set to 1. Only if the Structural Coefficient α is set to a value different from 1, theicon will be displayed in the Status Bar. Clicking on the icon brings up the Structural Coefficient Editor.
By default, the Node Cost of each node is set to one. If any of the nodes' cost is set to a value different from 1, the cost indicatorwill be displayed in the Status Bar. A click on this icon opens the Node Cost Editor.
BayesiaLab allows you to manage nodes in groups, or "classes". Some of these classes are predefined in their meaning, e.g. factors, but otherwise they can be arbitrarily created by the analyst. Whenever any kind of class is defined for the network, an indicator iconappears in the Status Bar.
By left-clicking the icon, you open open the Class Editor. By right-clicking, you bring up a list of all classes.
By default, all checkboxes are ticked as "visible". By unchecking, you can "tune out" a class from the current display, i.e. the corresponding nodes and arcs will appear as semi-transparent in the Graph Window. Also, if a class is unchecked, neither the nodes in that particular class, nor the arcs between these nodes, can be manipulated in the Graph Window.
The checkbox "All" allows you to quickly check/uncheck all checkboxes in the list. Checking "Intersection" allows you to easily define nodes that belong simultaneously to two or more clases.
Forbidden Arc Indicator
If Forbidden Arcs are defined for the current network, an iconis displayed in the status bar. Clicking on this icon opens the Forbidden Arc Editor.
Temporal Indices Indicator
The presence of Temporal Indices in the current network is indicated by theicon in the status bar. Clicking the icon opens the Temporal Indices Editor.
If Constants are defined in the network, theicon is displayed in the status bar. A click on the icon opens the Constants Editor. This icon is only accessible in the Modeling Mode.
Filtered States Indicator
If any node in the network contains Filtered States, theicon is visible in the Status Bar. Clicking on this icon prompts you to remove all Filtered States.
State Virtual Number Indicator
If any node has a virtual number of states, theicon is displayed in the status bar. A click on the icon opens the State Virtual Number Editor.
Local Structural Coefficient Indicator
If a Local Structural Coefficient is defined for any of the nodes, theicon appears in the status bar. A click on the icon opens the Local Structural Coefficient Editor.
Theicon in the Status Bar signals that Experts are associated with the current network. A click on the icon opens the Expert Editor for the network. By right-clicking on the icon, you can select to display the Assessment Report.
Time Variable Indicator
Within the Node Editor, the Probability Distributions of nodes can be defined by Equations.
If any of the Equations contain time as a variable, which would be denoted by "?t?", the "hourglass" icon will appear in the Status Bar.
Clicking on the icon will prompt you to remove the time variable from the network. However, this may have the consequence that the equations in the Node Editor are no longer valid and that the corresponding probability distributions are no longer defined. The warning icon will appear alongside the affected nodes.
Also, the icon can only be clicked in the Modeling Mode. In Validation Mode it is inactive and greyed out.
Evidence Scenario File
Whenever a Evidence Scenario File is available, theicon appears in the Status Bar. You can remove the current Evidence Scenario File by left-clicking on the icon.
If you hover with the mouse pointer over the icon, the number of available evidence sets in the Evidence Scenario File will be displayed.
Within Validation Mode, right-clicking the icon brings up the list of available evidence sets (enumerated by an index). You can now select (or "recall") any of the listed evidence scenario sets and thus apply them as evidence to the network. Any available comments will be shown in the Status Bar.
While you would typically learn a Bayesian network from data "from scratch", you can also utilize any available a-priori structural knowledge of the domain. This means that you can start with a manually constructed network that represents your existing knowledge and then refine this network by learning from data. However, this requires to "weight" the a-priori knowledge and the to-be-learned knowledge from data. As the a-priori knowledge is not associated with any underlying database, we need to arbitrarily specify how many cases/observations the a-priori knowledge should represent. This way we define a "Virtual Database" with N samples, which corresponds to the joint probability distribution of the a-priori network. Together with the "actual" database of M observations, BayesiaLab can then learn a new network from the N+M cases. Theicon indicates that such a Virtual Database is in use.
By clicking on the icon, you can remove the association between the network and the Virtual Database.
The Associated Database iconis perhaps the one symbol that appears more often than all others. It appears whenever data, i.e. a database, is associated with a network.
- If the Associated Database contains any missing values, a small question mark is added to the database icon:
- If the database contains a row/case weight variable, a symbol is added to the database icon:
- To indicate a learning/test sample split, a small "T" appears with the database icon:
- If the database is stratified, another symbol is added:
Altogether, the database icon can look like this:
Holding the mouse pointer over the database icon brings up a tooltip:
This tooltip will summarize a wide range of database attributes:
- the path and name of the database or, alternatively, that it is an internal database,
- the total number of cases/observations/rows/samples/examples in the database
- the number of cases in the learning set
- the number of cases in the test set
- the sum of row weights
- the stratification over a node and the corresponding probability distribution
- the presence of missing values.
By right-clicking the icon you can
- remove the data type, i.e. the split of the learning and test sample
- remove the database stratification
- remove the row/case weights
- display the database report
By left-clicking on the database icon you can entirely remove the association between the network and the database.
If the database has row identifiers, e.g. a survey ID or a case number, you can search for a specific row. Holding SHIFT plus right-clicking on the database icon brings up a search dialogue. This allows you to perform a full-text search on the row identifier field. For the search, you may use the wildcard characters ? and *. BayesiaLab follows the convention that ? stands for a single character, * may represent any set of characters including a null string. Upon entering the search criteria, the matching results are shown in list format, with the total number of matches displayed at the bottom.
Among this result set, you may now select any row by clicking on it. This will retrieve the observed values corresponding to the selected row identifier and set these values as evidence.
On occasion, you may wish to "lock" a network into Validation Mode, e.g. when sharing your network with others in order to prevent modification. Locking the network is available from the Network menu:
You will be prompted to enter a password for subsequent unlocking.
When the network is locked, theicon is displayed in the status bar. You can click on the icon and then enter the previously specified password to unlock the network. Upon unlocking, the icon is displayed in the status bar.
If previously locked network is currently unlocked, any click on the icon will lock it again immediately. The lock manager, which can be accessed from the Network menu, allows adding a lock, removing a lock, changing the password of the lock, etc.