#moon-causal

#Core capabilities

• Data preparation: rectangular dataset validation, data contracts, table operations, feature engineering, scaling, transformations, missing-value handling, and train/validation splits.
• Modeling: standardized logistic propensity regression, linear outcome regression, design matrices, polynomial terms, and treatment interactions.
• Estimation: IPW, stabilized IPW, ATT, ATC, overlap weighting, g-computation, doubly robust estimation, stratification, and nearest-neighbor matching.
• Diagnostics: standardized mean differences, variance ratios, overlap and positivity checks, calibration, effective sample size, residual diagnostics, and quality reports.
• Extended methods: subgroup effects, policy value, multi-arm treatments, continuous-dose response, difference-in-differences, event studies, synthetic controls, mediation, panel methods, time-series analysis, survey weighting, network interference, and survival curves.
• Production workflows: estimand registries, experiment design and power analysis, operational monitoring, fairness diagnostics, privacy-safe aggregates, robust regression, report serialization, and reproducibility manifests.
• Uncertainty and evaluation: bootstrap intervals, permutation inference, jackknife influence values, cross-validation, classification metrics, and deterministic synthetic benchmarks.

#Quick start

moon check --deny-warn
moon test --deny-warn
moon run cmd/main

#CLI

moon run cmd/main

moon-causal deterministic benchmark
sample_size=200
true_ate=1.75
estimated_ate=1.695501947510638
absolute_error=0.0544980524893619
standard_error=0.15562533888456542
effective_sample_size=177.61530018366125
propensity_auc=0.6822682268226823

#Architecture

• core_types.mbt, validation.mbt, and preprocessing.mbt define data contracts and transformations.
• linear_algebra.mbt, models.mbt, formula.mbt, and metrics.mbt provide numerical foundations and predictive models.
• estimators.mbt, matching.mbt, causal_methods.mbt, and advanced_causal.mbt implement treatment-effect estimators.
• balance.mbt, diagnostics.mbt, audit.mbt, and sensitivity.mbt provide diagnostics and robustness analysis.
• heterogeneity.mbt, multi_arm.mbt, continuous.mbt, panel.mbt, survival.mbt, and mediation.mbt cover extended designs.
• uncertainty.mbt, bootstrap_advanced.mbt, cross_validation.mbt, and simulation.mbt provide inference and reproducible evaluation.
• data_contracts.mbt, data_quality.mbt, feature_engineering_advanced.mbt, and table_ops_advanced.mbt support production data preparation.
• experiment_design.mbt, causal_pipeline_advanced.mbt, monitoring_reports.mbt, fairness_diagnostics.mbt, privacy_safe.mbt, and serialization_reports.mbt connect analysis to reviewable workflows.
• panel_advanced.mbt, time_series_advanced.mbt, survey_causal_advanced.mbt, network_interference.mbt, and survival_advanced.mbt provide specialized longitudinal and population-analysis methods.
• cmd/main contains a runnable benchmark application.

#Benchmark and reproducibility

• sample size and configured ground-truth ATE;
• estimated IPW ATE and absolute error;
• standard error and effective sample size;
• propensity-model AUC.

#Tests and quality checks

moon fmt --check
moon info
moon check --deny-warn
moon test --deny-warn
moon build --target wasm-gc
moon build --target native

#CI

• format and warning-as-error checks;
• wasm-gc and native builds;
• public-interface regeneration and clean-diff verification;
• wasm-gc tests with coverage summary;
• native type checks and tests.

#License

#Project links

• GitHub
• Gitlink

///|
test "README example" {
  inspect(mean([1.0, 3.0, 5.0]), content="3")
}