Entering observations (evidence) is one of the basic operations on a probabilistic model. It amounts to adjusting the model to a new situation, one in which more information is available. It allows to query the system subsequently about the new, posterior probability distributions. If you have gone through Building a Qualitative Bayesian network, you might remember that you have already entered evidence for the Hot Weather and Climate Control in the Building nodes. Let us go through this again.
You may load the model Productivity.qdsl created in Building a Qualitative Bayesian network from the Example Networks folder.
To enter evidence into your model, right-click on the node in question (in the picture below, node Climate Control in the Building) and choose Set Evidence.
Alternatively, select the node in question and choose Set Evidence submenu from the Node Menu.
Yet another way of entering evidence is double-clicking on one of the small square icons (
) in the lower-right corner of a node. In the image below, we double-clicked on the icon False (
) in both Climate Control in the Building and Hot Weather nodes.
Finally, perhaps the simplest way of entering evidence is pressing the letter T or F (for True and False respectively) once the node has been selected.
To retract evidence, double-click on the bold icon in a node with previously entered evidence (
) shown in the image above, press the letter C (for Clear) once the node has been selected, or right-click on the node and choose Clear evidence.
Alternatively, select the node in question and choose Clear evidence from the Node Menu.
You can also retract all evidence by choosing Clear all evidence from the Network menu:
To enter different evidence instead of retracting evidence altogether, just double click on a different state icon () or set different evidence from the Set Evidence sub-menu.