How can I specify statistic and attribute probes? And how can I plot confidence intervals and compare one statistic to another? (releases 8.x and 9.x)

Categories:
Solution Number:
S20759
Last Modified:
2013-08-20
Issue

How can I specify statistic and attribute probes? And how can I plot confidence intervals and compare one statistic to another? (releases 8.x and 9.x)

Solution

If you are using OPNET 10.0 or higher, please refer to FAQ#1395.To plot the confidence interval you will have to run a simulation with different seeds. Also you may want to obtain these results for multiple values of an attribute. The workflow is to:1. specify scalar probe for a statistic 2. specify attribute probes3. specify output scalar file4. Run simulations for different seeds and different values of the interested attribute5. Plot the scalar output of a statistic against the attribute values along with confidence intervalSetting a statistic probe:--------------------------1. Choose the statistic of interest: right click in the project workspace, select Choose Individual Statistics, choose the statistic of interest and close the dialog box2. Open the Probe editor, from the project editor menu Simulation => Choose Statistics (Advanced) 3. In the probe editor, right click on the statistic to be probed and edit attributes4. Set the value of attribute scalar data to enabled5. Choose the scalar type (eg: max_value, sample_mean etc..), this determines how the values obtained during a simulation run get represented by a single value.Setting a attribute probe:--------------------------1. Open the Probe editor (Simulation => Choose Statistics (Advanced)), if not open already2. In the probe editor create an attribute probe (Objects => Create Attribute Probe)3. Scroll down, and expand the Attribute Probes tree (click on the plus (+) sign)4. The newly created probe can be identified by no value under Attribute and Object columns5. Right click on the newly created probe, and select Choose Attributed Object6. Select the object of interest and click ok7. Right click again on the attribute probe and select edit attribtes8. Specify the attribute you probe (NOTE: The attribute you want to probe should be of a top level attribute and should be of type integer or double)Once you have specified the scalar probes save and close the probe editor.Specify a output scalar file:----------------------------1. Open the Simulation Sequence editor, from the project editor (Simulation => Configure Simulation (Advanced)). 2. Right click on the scenario icon and specify a file name in the Scalar file: text box.3. Save settings and close the Simulation Sequence editor.You will now have to run the simulation for different values of the probed attribute and different seeds. (For doing this you can duplicate this scenario and set different values to the attribute of interest in different scenarios. To Run all the scenarios from the project editor go to Scenarios => Manage Scenarios. Set the results column for each scenario either to <collect> or <recollect>. On clicking Ok all the scenarios will run). To plot the statistic probe against the attribute probe (and to plot the confidence intervals)1. You will have to run each scenario with multiple seeds 2. Open Analysis Configuration Editor (New => Analysis Configuration)3. Load scalar file you specified in the advanced simualtion configuration (File => Load Output Scalar File.. => choose the file name)4. Click on the second action button (Create graph from two scalars)5. Choose the Attribute probe for horizontal axis and the statistic probe for the vertical axis.6. The graph shows up, if you have had multiple runs for an attribute, you will see multiple y-values for the same x-value. 7. Right click on the graph and choose edit graph properties.8. Check the Show Confidence Interval box and choose the desired Confidence Interval percentage.

Environment

DES Kernel->Statistics and Reports

Attachments
NOTICE: Riverbed® product names have changed. Please refer to the Product List for a complete list of product names.
Can't find an answer? Create a case