{
  "schema_version": 1,
  "last_updated": "2026-08-27",
  "tiers": {
    "tier_1_foundational": [
      {
        "command": "regress",
        "estimator_mode": "ols_classical",
        "reference_implementation": "statsmodels.regression.linear_model.OLS (standard cov)",
        "coef_rtol": 1e-6,
        "se_rtol": 1e-5,
        "validation_status": "reference_validated",
        "notes": "Validated against statsmodels OLS point estimates, standard errors, t-stats, p-values, R2, and F-statistic."
      },
      {
        "command": "regress",
        "estimator_mode": "ols_robust_hc1",
        "reference_implementation": "statsmodels.regression.linear_model.OLS (cov_type='HC1')",
        "coef_rtol": 1e-6,
        "se_rtol": 1e-5,
        "validation_status": "reference_validated",
        "notes": "Validated against statsmodels OLS with HC1 robust heteroskedasticity-consistent covariance."
      },
      {
        "command": "regress",
        "estimator_mode": "ols_clustered",
        "reference_implementation": "statsmodels.regression.linear_model.OLS (cov_type='cluster')",
        "coef_rtol": 1e-6,
        "se_rtol": 1e-5,
        "validation_status": "reference_validated",
        "notes": "Validated against statsmodels cluster-robust covariance with finite-sample group adjustment."
      },
      {
        "command": "predict",
        "estimator_mode": "linear_prediction_and_residuals",
        "reference_implementation": "statsmodels.regression.linear_model.RegressionResults.predict / resid",
        "coef_rtol": 1e-6,
        "se_rtol": 1e-5,
        "validation_status": "reference_validated",
        "notes": "Validated fitted values (xb) and residual calculation against statsmodels."
      },
      {
        "command": "estat",
        "estimator_mode": "vif_and_ic",
        "reference_implementation": "statsmodels.stats.outliers_influence.variance_inflation_factor",
        "coef_rtol": 1e-5,
        "se_rtol": 1e-5,
        "validation_status": "reference_validated",
        "notes": "Validated Variance Inflation Factors and information criteria (AIC/BIC)."
      },
      {
        "command": "logit",
        "estimator_mode": "logit_mle",
        "reference_implementation": "statsmodels.discrete.discrete_model.Logit",
        "coef_rtol": 1e-5,
        "se_rtol": 1e-4,
        "validation_status": "reference_validated",
        "notes": "Validated Logit maximum-likelihood coefficients, Hessian-based SEs, log-likelihood, and pseudo-R2."
      },
      {
        "command": "probit",
        "estimator_mode": "probit_mle",
        "reference_implementation": "statsmodels.discrete.discrete_model.Probit",
        "coef_rtol": 1e-5,
        "se_rtol": 1e-4,
        "validation_status": "reference_validated",
        "notes": "Validated Probit maximum-likelihood coefficients, SEs, log-likelihood, and pseudo-R2."
      },
      {
        "command": "poisson",
        "estimator_mode": "poisson_mle",
        "reference_implementation": "statsmodels.discrete.discrete_model.Poisson",
        "coef_rtol": 1e-5,
        "se_rtol": 1e-4,
        "validation_status": "reference_validated",
        "notes": "Validated Poisson log-linear MLE coefficients, SEs, log-likelihood, and deviance."
      },
      {
        "command": "qreg",
        "estimator_mode": "quantile_regression",
        "reference_implementation": "statsmodels.regression.quantile_regression.QuantReg",
        "coef_rtol": 1e-5,
        "se_rtol": 1e-4,
        "validation_status": "reference_validated",
        "notes": "Validated median (q=0.5) and custom quantile estimates against statsmodels QuantReg."
      }
    ],
    "tier_2_econometric": [
      {
        "command": "ivregress",
        "estimator_mode": "2sls_and_gmm",
        "reference_implementation": "linearmodels.iv.model.IV2SLS / IVGMM",
        "validation_status": "implemented",
        "notes": "Unit tested with endogeneity and weak-instrument diagnostic workflows."
      },
      {
        "command": "xtreg",
        "estimator_mode": "panel_fe_re",
        "reference_implementation": "linearmodels.panel.model.PanelOLS / RandomEffects",
        "validation_status": "implemented",
        "notes": "Unit tested with entity/time fixed effects and random effects models."
      },
      {
        "command": "cfregress",
        "estimator_mode": "control_function",
        "reference_implementation": "statsmodels 2-step OLS control function",
        "validation_status": "implemented",
        "notes": "Unit tested with first-stage residual projection and bootstrap/2-step standard errors."
      },
      {
        "command": "did",
        "estimator_mode": "difference_in_differences",
        "reference_implementation": "statsmodels TWFE difference-in-differences",
        "validation_status": "implemented",
        "notes": "Unit tested with 2x2 and multi-period treatment setups."
      },
      {
        "command": "xtabond",
        "estimator_mode": "arellano_bond_gmm",
        "reference_implementation": "linearmodels dynamic panel GMM",
        "validation_status": "implemented",
        "notes": "Unit tested with lagged instrument moment conditions."
      },
      {
        "command": "xtlogit",
        "estimator_mode": "panel_logit_fe_re",
        "reference_implementation": "statsmodels conditional logit",
        "validation_status": "implemented",
        "notes": "Unit tested with panel binary choice workflows."
      }
    ],
    "tier_3_advanced": [
      {
        "command": "tobit",
        "estimator_mode": "censored_regression",
        "reference_implementation": "statsmodels / scipy.optimize MLE",
        "validation_status": "implemented",
        "notes": "Unit tested with left/right censoring boundaries."
      },
      {
        "command": "heckman",
        "estimator_mode": "sample_selection",
        "reference_implementation": "2-step Heckman selection model",
        "validation_status": "implemented",
        "notes": "Unit tested with inverse Mills ratio projection."
      },
      {
        "command": "zip / zinb",
        "estimator_mode": "zero_inflated_counts",
        "reference_implementation": "statsmodels ZeroInflatedPoisson / ZeroInflatedNegativeBinomialP",
        "validation_status": "implemented",
        "notes": "Unit tested with count and inflation submodel specs."
      },
      {
        "command": "streg",
        "estimator_mode": "survival_analysis",
        "reference_implementation": "statsmodels PHReg / lifelines",
        "validation_status": "implemented",
        "notes": "Unit tested with Weibull/Exponential/Cox proportional hazards."
      },
      {
        "command": "dml / drdid",
        "estimator_mode": "double_machine_learning_causal",
        "reference_implementation": "cross-fitting orthogonalized score estimators",
        "validation_status": "implemented",
        "notes": "Unit tested with K-fold cross-fitting and AIPW/doubly robust scores."
      },
      {
        "command": "lasso / ridge / elasticnet",
        "estimator_mode": "regularized_linear_models",
        "reference_implementation": "scikit-learn linear_model / statsmodels OLS.fit_regularized",
        "validation_status": "implemented",
        "notes": "Unit tested with cross-validation tuning loops (cvlasso, cvridge, cvelasticnet)."
      },
      {
        "command": "bayes",
        "estimator_mode": "mcmc_bayesian_regression",
        "reference_implementation": "bambi / pymc / arviz",
        "validation_status": "implemented",
        "notes": "Unit tested with multi-chain NUTS sampling, Gelman-Rubin diagnostics, and posterior intervals."
      },
      {
        "command": "spregress",
        "estimator_mode": "spatial_econometrics",
        "reference_implementation": "spreg / libpysal spatial lag and spatial error models",
        "validation_status": "implemented",
        "notes": "Unit tested with Queen/Rook/KNN/GAL/GWT weights matrices."
      }
    ]
  }
}