The Policy Summary chart display the probability distributions for the events in a model given that the optimal policy is followed (referred to as policy-dependent probabilities). This is useful for answering questions such as "If I follow this policy, what is the probability that this event outcome will occur?" Policy Summaries can be generated during a Decision Analysis or from a Policy Tree (Policy | Policy Summary | Add when looking at a Policy Tree). If you would like to see policy-dependent probabilities for all events in the model, specify Number of policy levels to save to be the maximum when you run a decision analysis.
Understanding Policy Summaries
In a Policy Summary, DPL displays policy-dependent (or policy-conditional) probability distributions for each event for which policy data was gathered. For each event, DPL displays the event's states plus an additional state called "(does not occur)". The probability indicated for a state is the probability of the state conditioned on the optimal policy.
The (does not occur) state indicates the probability that the event is not encountered under the optimal policy. This may happen in asymmetric trees, where the same events do not occur on all paths through the tree.
For chance nodes whose probabilities are not conditional, the probabilities of each state occurring will be the same as those entered in node data unless the event is not encountered in some paths through the tree.
Downstream Decisions in Policy Summaries
Note that the decisions in Policy Summaries have probabilities associated with their alternatives. This is the probability that this choice will be made if the optimal policy is followed. When a decision occurs at the beginning of the tree the probability of the optimal choice is 1.0 and the probabilities for the other alternatives are 0.0. However, when a decision follows one or more chance event, the optimal choice for the decision may depend on the outcome of those events, in which case the probability for each decision alternative may be less than 1.0.
The Policy Summary is particularly useful when a decision occurs fairly late in a large tree. The decision may appear several times in the policy, making it hard to quickly see whether the decision depends on the preceding events or if there is always (or usually) one dominant alternative. With the Policy Summary, you can quickly tell to what extent downstream decisions are dependent on the preceding events.
Formatting Policy Summaries
You have the same formatting options in the Policy Summary window that you have in the decision tree: you can move and align nodes, add text blocks, change the fonts, print, and copy to the Clipboard.
Versions: DPL Professional, DPL Enterprise, DPL Portfolio