The Run Settings dialog stores settings for running a DPL analysis. It can be accessed by clicking the dialog box launcher for Home | Run or Home | Sensitivity. Some of these settings can also be changed directly on the ribbon or as you run a sensitivity from those run dialogs.
Samples for Monte Carlo and discrete tree simulation
The number of paths to evaluate in a Discrete Tree Simulation or the number of initial samples to draw for a Monte Carlo simulation. A larger number improves accuracy but increases run-time. The number must be between 1 and 2,000,000,000.
Minimum at decisions
If a model contains a downstream decision (i.e., a decision that occurs after one or more initial uncertainties have been resolved), then this parameter controls the minimum number of samples to draw for each alternative at the downstream decision. A larger number improves accuracy but increases run-time. The number must be between 1 and 2,000,000,000.
Compile time optimizations
The types of optimization are Expression Optimization, Series Optimization, and Lottery Optimization. In general, these optimization methods increase memory (RAM) requirements and reduce run-times. If you are running out of memory, you should turn these options off. Lottery optimization is off by default because it is the most memory intensive of the three. These options do not affect the numeric results. NOTE: The effect of these options are highly model-dependent and may lead to an increase in run times. You should experiment with these options to determine the optimal setting for your model. If your model is linked to a spreadsheet, remember that the majority of the run-time is dictated by the calculations taking place in Excel.
Optimizes the way large expressions are calculated. Examples of large expressions are get/pay or node data expressions involving a number of variables and/or advanced formulas. If you do not have such expressions in your model this optimization method will have no impact on your run time.
Optimizes the way series expressions are handled. If you do not have series expressions in your model this optimization method will have no impact on your run time.
Optimizes the method by which the lottery is calculated. This optimization method has the greatest impact of the three optimization methods and will affect the run time of most models.
Number of intervals sets the maximum number of intervals that are used for generating Risk Profiles. A small number gives a relatively coarse distribution, whereas a large number (maximum of 1500) gives a relatively fine distribution. If the number specified exceeds the number of actual endpoints, then the actual number of endpoints is used. Reducing the number of intervals reduces run time and memory usage.
Increasing the number of intervals has no impact on the calculation of expected value for a model. However, a larger number of intervals does result in more accurate percentiles. If you are interested in percentiles for a risk profile or are running a Time Series Percentiles, you may wish to increase this.
For models with continuous chance nodes and no decisions, you may select an all samples distribution.
If you have recorded endpoints, you may run a Scenario Risk Profile from any point in the Policy Tree. When you do so, you will prompted for the states of each event that you want included in the set of scenarios for the Scenario Risk Profile. If you do not wish to be prompted and instead want to automatically use the selection from the Policy Tree, check this box.
For Scenario Risk Profiles, check the Use selection from Policy Tree box if you do not want to be prompted for which scenarios to include in the Scenario Risk Profile.
Within the Default Type section, you can specify whether the graph should initially be shown in cumulative ("S-Curve") form or as a frequency histogram (bar chart). This doesn't affect the data generated, and if you change your mind after the run you can change graph types in Chart | Format | Display.
Clear unsaved outputs without prompting
Check this box if you don't want to be prompted each time you delete and unsaved output.
Large number of endpoint display
When you record a large number of endpoints, by default DPL will prompt you as to whether you want them displayed in a grid since the grid is memory intensive. If you'd prefer to always display or never display, select one of these options. This option is available only if you access the Run Settings dialog via Home | Run | Options. You can specify what should be defined as a "large number" of endpoints within the text box provided. By default it is set to 100,000 endpoints.
Check the Gather attribute expected values in the Policy Tree button to have DPL roll back and display the expected values of all attributes. This option is available only if the model has multiple attributes and the Run Settings dialog was accessed via Home | Run | Options. You can also access this option via the Objective & Utility dialog.
Rainbow Diagram steps
By default DPL will warn you if the number of steps for a rainbow diagram run is large. De-select the box to turn off the warning. This option is available only if you access the Run Settings dialog via Home | Sensitivity | Options.
Default States for Option Value Charts
For Option Value Charts, check the Use default states defined in node definition for Option Value Chart if you do not want to be prompted for which alternative to control each decision. This option is available only if you access the Run Settings dialog via Home | Sensitivity | Options.
Parallel Endpoint Recording (DPL Enterprise and DPL Portfolio)
For each machine (client and server) that will perform parallel endpoint recording specify the working directory that contains the DPL model and spreadsheet to be analyzed in the box provided. For the server computer(s), click the Act as Server button. DPL will then indicate that it's waiting for a command.
Versions: DPL Professional, DPL Enterprise, DPL Portfolio