How to Perform MultiRun Simulations

Use the existing SimPaths multi-run entry point rather than creating a separate JAS-mine demo wrapper. It runs experiments sequentially.

Define the experiment

Decide whether you need repeated draws, different policy scenarios or different model assumptions. These are different designs; Uncertainty Analysis explains how to interpret them.

Configure and run

1. Follow Multiple Runs for setup and the complete configuration reference.
2. Copy config/default.yml to a named experiment file under config/.
3. Set the run count, initial seed, start/end years, population size and output settings.
4. Enable only the intended innovations. randomSeedInnov increments the seed between runs; it does not isolate simulation randomness when coefficient resampling is also enabled.

5. Run the configuration from the repository root. For a file named experiment.yml:

   java -jar multirun.jar -config experiment.yml
   

6. Check the logs, run count, labels and output files before launching a larger experiment.

Keep an unchanged copy of the configuration, policy schedule and input versions with the results. For a set of policy scenarios, use separately named configurations and an explicit record of their differences.

Extend the run design

If the existing controls cannot express the design, inspect buildExperiment(), nextModel(), iterateParameters() and setupRunLabel() in MultiRun Implementation. Changing a YAML key alone does not add a new iteration rule.

The JAS-mine MultiRun architecture explains the library contract; it is not another set of SimPaths run commands.