Papilon
Copyright 2024-2026 Brian Curry / Vector1 Research

This product includes software developed by Brian Curry / Vector1 Research
(https://vector1.ai).

"Papilon", "PyCausalSim", "Vector1" and "Vector1 Research" are trademarks of Brian Curry.
The Apache License, Version 2.0 does not grant permission to use these names
or marks, except as required to describe the origin of the work (see Section 6
of the License and TRADEMARKS.md).

Papilon depends on, but does not include, NumPy, SciPy, pandas, scikit-learn,
NetworkX, Matplotlib and seaborn, each distributed under its own license.

Methods first published in this repository, with the release that introduced them
(see CHANGELOG.md and docs/design/ for dated descriptions):

  - Gated importance sampling for marketing mix models, with Bayes-factor
    selection of carryover and saturation families (0.3.0, 0.5.0)
  - Downside-protected budget planning with support flags (0.4.0)
  - Self-check on simulated copies of a brand, lift-test value by posterior
    reweighting, and the MarketingGraph ontology with its audit (0.5.0)
  - The learning loop with value-of-information experiment choice and budget-neutral
    exploration between indistinguishable channels, graded against known truth;
    conjugate geo pooling with sampled shared curves (1.0.0)

The marketing vocabulary at https://vector1.ai/ns/marketing# is maintained by
Vector1 Research. Derivative works must keep this NOTICE (Apache License 2.0,
Section 4(d)) and must not imply endorsement by, or affiliation with, Brian Curry
or Vector1 Research.
