Examples

Pick the guide that matches how treatment is represented in your data.

Start from the shape of t. The guide changes with the treatment representation, but the estimator contract stays the same.

The Core Workflow Does Not Change

The same pattern shows up in every tutorial.

  1. Build X, t, and y from observed covariates, treatments, and outcomes.
  2. Fit one estimator that matches the treatment type and estimand.
  3. Create the requested intervention table or treatment grid.
  4. Predict and compare the returned causal quantity.

Choose a path

Start With The Treatment Shape

The examples use synthetic datasets to compare with ground truth, but the workflow is the same for real data.

If treatment looks like Open Main output
One numeric column such as dose, score, or exposure Continuous treatments A response curve on a treatment grid.
One discrete column with named levels such as control, placebo, and treated Categorical treatments Average potential outcomes for requested levels.
Multiple treatment columns such as dose plus regime Multidimensional treatments Joint intervention comparisons across treatment regimes.

Shared workflow