Using an Endpoint Database
If you have previously recorded an Endpoint Database while doing a decision analysis it can be used to generate additional outputs of a decision analysis even if you have modified your model since recording the Endpoint Database. This can save time significant run time if your model is large or is linked to a large spreadsheet. If an Endpoint Database is available for playing, the Full Tree Enumeration from Endpoints item in the Home | Run | Decision Analysis drop-down will be enabled.
When you play an Endpoint Database, DPL traverses the decision tree and when it reaches an endpoint it looks up the objective function value (and any attribute values if present) for the endpoint at the end of the tree rather than recalculating these values. In addition, DPL ignores any get/pay expressions in the tree since it knows it will be looking up endpoint values rather than calculating them. DPL uses the probability data from the Influence Diagram to roll-back the decision tree with the looked-up endpoint values.
When Can the Endpoint Database be Re-Used
You may re-use an endpoint database if you have re-ordered your decision tree, controlled branches, changed probability data or introduced asymmetries to the decision tree. You may also re-use the Endpoint Database if you have added events to your model as long as they do not directly change the value model. E.g., you may add a chance node that probabilistically conditions another chance node.
When Can the Endpoint Database Not be Re-Used
If you change anything in your value model subsequent to recording endpoints, you may not re-use the endpoint database. E.g., if you add a chance node with values associated with it, you will not be able to re-use the Endpoint Database. Or if your model is linked to an Excel spreadsheet and you changed calculations in it for NPV or other values that DPL uses, then you will not be able to re-use the Endpoint Database.
Before playing endpoints, DPL checks to see if they are valid. If DPL finds that your value model has changed since you recorded endpoints it will issue a warning and you will not be allowed to play the endpoints.
In certain situations even if you do not change your value model, you may not be able to re-use the Endpoint Database. For example, this could occur when you record an Endpoint Database for an asymmetric tree. If you subsequently re-order events in the tree and try to play endpoints, you may not be able to. This is because DPL may need to look up an endpoint that does not exist due to the asymmetries in the initial tree when the endpoints were recorded. In this situation, DPL will issue a warning indicating that an endpoint was not found and it will abort the run.
Versions: DPL Professional, DPL Enterprise, DPL Portfolio