Imposed policies as findings derives to invalid networks.

[Jorge Pérez]

Related to Issue #351. In a network with decisions, if we impose a policy (Do test? = yes in the attached network) and move to inference mode (propagation), the policy is showed as a finding.

The critical issue is that, when we return to edition mode, the decision still have a finding assigned.

Recently, we decided to remove the option of adding/editing/removing findings from decision nodes, so it is necessary to re-open the network wihout saving it to remove that findings. If we save the network the parser put that evidence in the decision node (without warning the user) and it is impossible to remove the findigns without removing from the pgmx code.

[Manuel Luque]

The network attached above (D-decide-test.pgmx) is even corrupted by the bug itself, as it also contains two evidence cases in the PGMX file:

<Evidence><EvidenceCase><Finding variable="Do test?" state="yes" /></EvidenceCase><EvidenceCase><Finding variable="Do test?" state="yes" /></EvidenceCase></Evidence>

If we want to reproduce this issue, we have to use a network without that evidence. I attach here a the ID-decide-test network by removing the evidence cases from the PGMX file.

[Manuel Luque]

This issue is fixed for the attached network, which is an influence diagram. I need more insight about problems in Markov influence diagrams (I invite people to create related issues).