Pillars

ergodic is organized around three pillars. Each one reads the shared foundations, a graph and a typed dataset, and answers a different question.

• Causal discovery

  ---

  Learn causal structure from data. Constraint-based searches (PC, FCI, RFCI), score-based ones (GES, BOSS, GRaSP), and the functional DirectLiNGAM, with time series through PCMCI.

  Causal discovery

• Causal inference

  ---

  Estimate causal effects from a graph and data. A library of estimators from plain adjustment to double machine learning, with per-unit effects and an optional Bayesian posterior.

  Causal inference

• Process intelligence

  ---

  Discover and analyze processes from event logs: a process map, a causal case table, a mined process model, and decision-point analysis.

  Process intelligence

When the rollout is the experiment and no graph carries it, the panel designs sit beside the inference pillar: difference-in-differences, synthetic control, and interrupted time series.