The evaluation method controls the evaluation of the model during a DPL analysis. When you run a decision analysis by clicking on the Decision Analysis button, DPL uses the default or last used evaluation method to run the model which is indicated by the icon. You may also use the drop down on the Decision Analysis split-button to explicitly select which evaluation method to use. The default or last used method will also be used when you run sensitivities (e.g., tornado or rainbow diagrams) on the model. You can set which evaluation method will be default and used for sensitivities within the (Sensitivity) Run Settings dialog. This can be accessed by clicking the dialog box launcher within the Home | Run or Home | Sensitivitygroups.
Fast Sequence Evaluation
Allows DPL to take advantage of special structural properties of your model to reduce computation time. This method produces an exact calculation for expected value and a close approximation for the full Risk Profile. It is usually much more computationally efficient than full tree enumeration. It is the default method for models not linked to a spreadsheet.
Full Tree Enumeration
Tells DPL to evaluate each path in the decision tree regardless of whether structural properties of your model may be able to reduce run-time. This will typically increase run-time for most models. Full Tree Enumeration may be the only evaluation method available for models that include non-linear objective, constraint, or user-defined utility functions.
If you check Endpoints in the Home | Run group, then Full Tree Enumeration will be used since it is required in order to record endpoints. Full Tree Enumeration also produces the "smoothest" Risk Profile. It is the default method for models linked to a spreadsheet.
Discrete Tree Simulation
Tells DPL to evaluate only a randomly selected subset of the decision tree paths based on a number of samples that you specify. This method decreases run-time. It produces an approximate expected value and Risk Profile percentiles. To set the Discrete Tree Simulation Options click the dialog box launcher for Home | Run, set the Evaluation Method to Discrete tree simulation, and click the Settings button.
Monte Carlo Simulation
To run a Monte Carlo simulation, you must have at least one continuous chance node in your model. If you have one or more continuous chance nodes, it is the only evaluation method available. During a run, the continuous chance node(s) are simulated by drawing random samples from named distributions with specified parameters. It produces an approximate expected value and Risk Profile percentiles. Be sure to choose the appropriate number of Initial samples and Restart samples from the drop down box under Monte Carlo Samples within the Home | Run group prior to running a Monte Carlo Simulation. Please see Running a Monte Carlo Ssimulation for a description of Monte Carlo samples.
Full Tree Enumeration from Endpoints
Tells DPL to analyze the model by using a previously recorded Endpoint Database so that most expressions don't need to be recalculated. This evaluation method is only available if endpoints were recorded previously for the model. Risk Profile percentiles and expected values produced are exact. This method allows you to re-run the model quickly and can be quite time-saving, particularly if your model is linked to a sizable spreadsheet.
Record Endpoints using Servers
Tells DPL to record endpoints in parallel using one or more DPL endpoint servers. Once recorded the endpoint will be available for a run via Full Tree Enumeration from Endpoints. For more information on this evaluation method see Parallel Endpoint Recording.
Act as Server
Act as a server for parallel endpoint recording. This instance of DPL will wait for a request to record endpoints from another instance running on a different computer.
Versions: DPL Professional, DPL Enterprise, DPL Portfolio
See Also
Running a Decision Analysis