Regional extension

Purpose

The regional simulator reuses the complete baseline generation, validation, table-selection, and writer pipeline. It adds a behavior provider that may alter traffic scale, conversion log odds, and conversion-noise spread from regional context.

The repository retains regional reference configurations and generated outputs for Europe, Japan, and North America so the documented experiments remain reproducible. Their economic indicators and campaign mappings are simulation inputs with recorded confidence, not measured Amazon shopper behavior or actual brand campaign results. examples/regional.toy.json remains the small fictional fixture used by automated tests.

Configuration inheritance

A private regional file can inherit a private baseline file with extends. Relative paths are resolved from the child file itself, not from the installed package or current working directory.

{
  "extends": "../baseline/campaign.json",
  "regional_behavior": {
    "reference_internet_reach_rate": 0.8,
    "reference_target_audience_density": 0.1,
    "reference_income_inequality_gini": 0.35,
    "economic_willingness_log_odds_weight": 1.0,
    "income_inequality_noise_weight": 1.0
  }
}

These numbers demonstrate file shape only. They are not estimates for a real place.

Behavior transformation

Regional traffic scaling compares internet reach and target-audience density with explicit reference values. Economic willingness adjusts conversion log odds. The income-inequality field changes noise spread, not average demand.

$$\begin{array}{rl}\text{inequality difference}& =\text{income inequality Gini}-\text{reference income inequality Gini}\\ \text{noise multiplier}& =max(0,1+\text{income inequality noise weight}\times \text{inequality difference})\\ \text{regional conversion noise spread}& =\text{baseline conversion noise spread}\times \text{noise multiplier}\end{array}$$

The maximum bound prevents an invalid negative standard deviation.

Run with a private file

mta-regional-dataset \
  --config /private/project/configs/region.json \
  --output /private/project/generated/region \
  --table path

The same --storage, --table, and --writer options are available as in the baseline command.

Evidence responsibility

A private caller should record source organization, indicator, observation year, retrieval date, transformation, uncertainty, and confidence for every real regional input. National statistics are not measurements of Amazon shoppers or campaign response.

See Regional Parameters and Recommendations.