Advanced
Explicit experimental designs, population assumptions and saved scientific records. The same engine and validation as the original builder.
Understanding the scientific settings
These settings declare assumptions. They are not empirically fitted estimates.
A parameter distribution assigns different input values across synthetic agents.
Correlated sampling allows declared parameters to vary together.
A subgroup applies declared settings to part of a synthetic population.
A sensitivity sweep repeats a design at different declared parameter values.
Model comparison runs alternative policies under declared conditions.
Replications repeat a configured experiment with separately derived random streams.
Paired seeds reuse matched random draws across compared conditions; independent seeds draw separately.
Softmax allows some randomness in choices: options with higher model value are more likely, but not certain, to be chosen.
An observed share is the fraction of agents choosing an option in one saved simulation run.
Provenance records the configuration and software identity associated with an execution.