- Home
- Skills
- Research & Analysis
- Causal ML Complete: 9 Skills
More screenshots
Works with the AI Agents you already use
Causal ML Complete: 9 Skills
All nine causal research skills: theory, estimator selection, DML, executable code, reproducible runs, heterogeneity, auditing and reporting. Grounded in official EconML/DoubleML documentation.
$3.99
Causal ML Complete: 9 Skills
Example session with this skill installed
Use the bundled synthetic data to run the joint model with v, d, d squared and v times d. Residualize every declared basis separately, estimate all four coefficients together, and report the full output and inference boundary.
- Read your context and instructions
- Compiled the causal ml complete
Reproducible synthetic run
Executed the bundled demo with 1,800 synthetic observations, seed 71 and three cross-fitting folds. These are simulated data; the letters v and d are generic exposures.
The declared response basis is (v, d, d squared, v times d). Each basis column is separately predicted from the controls on training folds. Held-out residuals are then entered in one joint final regression.
| Term | Known coefficient | Estimate | SE | 95% CI |
|---|---|---|---|---|
| v | 0.60 | 0.596038 | 0.012107 | [0.572308, 0.619768] |
| d | -0.50 | -0.487961 | 0.011010 | [-0.509539, -0.466382] |
| d_squared | 0.35 | 0.354615 | 0.005942 | [0.342969, 0.366261] |
| v_x_d | 0.25 | 0.244946 | 0.009102 | [0.227107, 0.262785] |
Inference and files
The displayed standard errors use orthogonal-score HC0 with an asymptotic normal reference. This run is iid; it does not claim clustered uncertainty. Files include the frozen input/config/script hashes, package versions, fold-level counts, out-of-fold predictions, complete estimates, residual-basis VIF and final status.
Interpretation
The estimated coefficients recover the known simulation parameters closely. In this polynomial model, the derivative with respect to d is beta_d + 2 beta_d_squared times d + beta_v_x_d times v. It is not the same hypothesis as the uncentered linear coefficient.
Validation boundary
The package has 33 passing synthetic/numerical tests, including 12 official-library adapter smoke tests. Synthetic recovery checks implementation; a causal conclusion for real observations still requires a defensible design, overlap, appropriate dependence handling and nuisance-rate conditions.
Connects securely to your tools. The creator never sees your data.
About this skill
The complete research workflow
All nine modules: causal-theory, causal-method-router, cml-estimators, dml-estimation, causal-code, causal-run, causal-heterogeneity, causal-audit and causal-report.
Complete includes every Essentials module and adds guidance for CATE/GATE, explicit subgroup contrasts, support checks, hidden-confounding sensitivity, bad controls, dependence-aware inference and calibrated reporting. Heterogeneity guidance includes when a separate implementation is required; the CLI does not claim to execute every documented analysis.
What you receive
Original SKILL.md instructions, progressively loaded equations and references, three standalone Python scripts, a pinned dependency list and an explicit validation record. Sources are the official Microsoft EconML and DoubleML documentation.
Executable scope
- Native joint-basis PLR and binary ATE AIPW, including grouped folds and score covariance.
- DoubleML PLR, IRM, PLIV and IIVM.
- EconML S/T/X learners, LinearDML, LinearDRLearner, CausalForestDML, OrthoIV and DMLIV.
- Reproducibility manifests, out-of-fold results, declared seeds and failure checks.
The 33-test suite passed locally and in GitHub CI. Inference is estimator-specific. OrthoForest, DRIV and DeepIV are discussed in the references; their execution requires additional implementation. No API key is required for local estimation.
MIT source: https://github.com/Mat-Wong/Causal-ML-Skills . This listing provides a conveniently packaged distribution of the public source.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Research & Analysis.
Causal identification is not established by fitting. Only numeric inputs are supported by the bundled CLI. Official-library adapters use iid inference; native grouped score covariance is available separately. Some documented estimator families require further implementation. No significance or causal-validity guarantee.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 8 days ago
- Software
Python 3.12NumPypandasSciPyscikit-learnEconML 0.16.0DoubleML 0.10.1pytestPyYAML
EconML and DoubleML power the library adapters; pytest and PyYAML run the validation.
Permissions
Reads user-selected CSV/config files and writes a new analysis directory. Uses terminal execution and may install pinned dependencies with network access. No credential collection, data upload, scraping or paid model calls.