causal-tree-decomposition

Decompose complex predictions into causal trees. Compute combined probability, run sensitivity analysis and what-if scenarios. Zero dependencies.

import { CausalTree, node, or } from 'causal-tree-decomposition'

const tree = new CausalTree('Oil exceeds $100 by Dec 2026')
tree.add(or('Supply disruption', 0.3, [
  node('Hormuz closure', 0.15),
  node('Russia further cuts', 0.2),
]))
tree.add(node('OPEC cuts production', 0.25))

tree.compute()                        // 0.52
tree.sensitivity()                    // [{ label: 'OPEC cuts', impact: 0.12 }, ...]
tree.whatIf('Hormuz closure', 0.95)   // { probability: 0.81, delta: +0.29 }
console.log(tree.toString())

Install

npm install causal-tree-decomposition

API

• CausalTree(title, operator?) — create tree (default OR)
• .add(node) — add root node
• .compute() — combined probability
• .sensitivity() — which nodes matter most
• .whatIf(label, newProb) — counterfactual
• .toJSON() / fromJSON() — serialize
• .toString() — pretty print
• .flatten() — all leaf nodes

Helpers

• node(label, prob) — leaf node
• or(label, prob, children?) — OR node
• and(label, prob, children?) — AND node

License

MIT — SimpleFunctions