[Repo Assist] fix: raise ValueError for unsupported identifier_method in covariate adjustment estimators

[github-actions]

🤖 This PR was created by Repo Assist, an automated AI assistant.

Closes #618

Root Cause

Covariate-adjustment estimators (LinearRegressionEstimator, GeneralizedLinearModelEstimator, all PropensityScore* estimators, and DistanceMatchingEstimator) silently produced incorrect results when used with incompatible identification strategies such as frontdoor, iv, or mediation.

When identifier_method is "frontdoor", get_adjustment_set() falls back to returning empty backdoor variables, causing the estimator to run without the right covariates and yielding a biased, incorrect effect estimate — without any error or warning.

DoublyRobustEstimator already has this guard; the other covariate-adjustment estimators did not.

Fix

Add an early ValueError to the fit() methods of:

• RegressionEstimator (covers LinearRegressionEstimator and GeneralizedLinearModelEstimator)
• PropensityScoreEstimator (covers PropensityScoreMatchingEstimator, PropensityScoreWeightingEstimator, PropensityScoreStratificationEstimator)
• DistanceMatchingEstimator

The check fires before any computation and gives a clear message directing users to TwoStageRegressionEstimator (for frontdoor/mediation) or InstrumentalVariableEstimator (for iv).

TwoStageRegressionEstimator is unaffected because it already overrides identifier_method to "backdoor" on the internal estimand copies it passes to regression sub-estimators.

Tests

Added test_invalid_identifier_method_raises to TestLinearRegressionEstimator, parametrized over frontdoor, iv, and mediation.

Test Status

tests/causal_estimators/ — 31 passed in 591s

All estimator tests pass, including the existing two-stage regression (frontdoor) tests which confirm the internal identifier_method = "backdoor" override is unaffected by this change.

Note

🔒 Integrity filter blocked 55 items

The following items were blocked because they don't meet the GitHub integrity level.

• ImportError: weighting_sampler is not an existing do sampler. #71 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• DoSampler: Fails when you Add a Graph to the Causal Model #83 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Errors of Notebook: The Causal Story Behind Hotel Booking Cancellations #198 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Mediation Analysis key error #214 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• arrays used as indices must be of integer (or boolean) type #225 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• effect_estimate different using do operator and _estimate_effect function #309 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• AttributeError: 'CausalEstimate' object has no attribute '_estimator_object' #357 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• CausalModel.estimate_effect - UnboundLocalError: local variable 'identifier_name' referenced before assignment #365 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Error in Conditional Effect Estimation in backdoor.linear_regression_estimator #401 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• 'fit_estimator' argument in 'estimate_effect' function of 'CausalModel' object #414 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Problems with multi-valued treatment and parameter 'evaluate_effect_strength' #416 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Setting the seed for estimate effect methods #418 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Categorical Covariates in Forest models #471 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• How can i get a deterministic output/estimate? #556 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• dowhy.plotter not presenting in Azure ML #667 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• effect_modifiers #767 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• ... and 39 more items

To allow these resources, lower min-integrity in your GitHub frontmatter:

tools:
  github:
    min-integrity: approved  # merged | approved | unapproved | none

Generated by Repo Assist · ◷

To install this agentic workflow, run

gh aw add githubnext/agentics/workflows/repo-assist.md@b897c2f3e43bde9ff7923c8fa9211055b26e27cc

[github-actions]

Commit pushed: 62d434c

Generated by Repo Assist

[github-actions]

🤖 This is an automated update from Repo Assist.

CI fix: Pushed commit 3fa5c39 to address the notebook test failure (dowhy_functional_api.ipynb).

Root cause of the failure: The validation check used a whitelist (identifier_method not in ["backdoor", "general_adjustment"]). When using the functional API (identify_effect() rather than CausalModel.identify_effect()), identifier_method is never set and remains None. None is not in the whitelist, so the new guard was incorrectly raising ValueError.

Fix: Changed the check to identifier_method is not None and identifier_method not in [...]. This:

• Allows None through (functional API case — effectively backdoor-like)
• Still raises ValueError for explicit "frontdoor", "iv", "mediation" (the original intent)

Also added a regression test test_none_identifier_method_does_not_raise to TestLinearRegressionEstimator to prevent this from regressing.

Note

🔒 Integrity filter blocked 75 items

The following items were blocked because they don't meet the GitHub integrity level.

• #1445 list_pull_requests: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• #1444 list_pull_requests: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• #1418 list_pull_requests: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• #1399 list_pull_requests: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• #1396 list_pull_requests: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• #1392 list_pull_requests: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• #1371 list_pull_requests: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Unable to estimate causal effect with intermediary variable? #69 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• ImportError: weighting_sampler is not an existing do sampler. #71 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• DoSampler: Fails when you Add a Graph to the Causal Model #83 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Errors of Notebook: The Causal Story Behind Hotel Booking Cancellations #198 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Mediation Analysis key error #214 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• arrays used as indices must be of integer (or boolean) type #225 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• How the Identification Works and About Two Outcomes #240 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• How to get the control group with propensity score matching? #242 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Mediation analysis: how to choose mediators? #249 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• ... and 59 more items

To allow these resources, lower min-integrity in your GitHub frontmatter:

tools:
  github:
    min-integrity: approved  # merged | approved | unapproved | none

Generated by Repo Assist · ◷

To install this agentic workflow, run

gh aw add githubnext/agentics/workflows/repo-assist.md@b897c2f3e43bde9ff7923c8fa9211055b26e27cc

[amit-sharma]

good check to add