CLI: Graph¶

causal-toolkit graph --config DAG.yaml [OPTIONS]

Config Format (dag.yaml)¶

edges:
  - ["age", "treatment"]
  - ["income", "treatment"]
  - ["age", "outcome"]
  - ["income", "outcome"]
  - ["treatment", "outcome"]
treatment: "treatment"
outcome: "outcome"
common_causes: ["age", "income"]
instruments: []
mediators: []

Options¶

Option	Short	Description
`--render`	`-r`	`matplotlib`, `plotly`, `graphviz` (default: `matplotlib`)
`--backdoor/--no-backdoor`		Highlight backdoor paths (default: true)
`--out`	`-o`	Output file (without extension)

Examples¶

# Matplotlib PNG
causal-toolkit graph -c dag.yaml -r matplotlib -o mydag

# Interactive HTML
causal-toolkit graph -c dag.yaml -r plotly -o mydag

# Graphviz (high quality)
causal-toolkit graph -c dag.yaml -r graphviz -o mydag

Output¶

Found 2 backdoor path(s):
['treatment', 'age', 'outcome']
['treatment', 'income', 'outcome']

Minimal adjustment sets:
[['age', 'income']]

Do-calculus steps:
Target: P(Y|do(T))
Step 1: Backdoor criterion satisfied with adjustment sets:
  Set 1: {age, income}
Step 2: Apply backdoor adjustment: P(Y|do(T)) = Σ_{C∈Adj} P(Y|T,C)P(C)