User guide

• 1. The basics of double/debiased machine learning
  • 1.1. Data generating process
  • 1.2. Regularization bias in simple ML-approaches
  • 1.3. Overcoming regularization bias by orthogonalization
  • 1.4. Sample splitting to remove bias induced by overfitting
  • 1.5. Double/debiased machine learning
  • 1.6. Partialling out score
  • 1.7. References
• 2. The data-backend DoubleMLData
  • 2.1. DoubleMLData from dataframes
  • 2.2. DoubleMLData from arrays and matrices
• 3. Models
  • 3.1. Partially linear regression model (PLR)
  • 3.2. Partially linear IV regression model (PLIV)
  • 3.3. Interactive regression model (IRM)
  • 3.4. Interactive IV model (IIVM)
  • 3.5. Difference-in-Differences Models (DID)
• 4. Heterogeneous Treatment Effects
  • 4.1. Group average treatment effects (GATEs)
  • 4.2. Conditional average treatment effects (CATEs)
  • 4.3. Weighted Average Treatment Effects
  • 4.4. Quantiles
  • 4.5. Conditional value at risk (CVaR)
  • 4.6. Policy Learning with Trees
• 5. Score functions
  • 5.1. Implementation of the score function and the estimate of the causal parameter
  • 5.2. Implemented Neyman orthogonal score functions
  • 5.3. Specifying alternative score functions via callables
• 6. Double machine learning algorithms
  • 6.1. Algorithm DML1
  • 6.2. Algorithm DML2
  • 6.3. Implementation of the double machine learning algorithms
• 7. Learners, hyperparameters and hyperparameter tuning
  • 7.1. Python: Learners and hyperparameters
  • 7.2. R: Learners and hyperparameters
• 8. Variance estimation and confidence intervals
  • 8.1. Variance estimation
  • 8.2. Confidence bands and multiplier bootstrap for valid simultaneous inference
  • 8.3. References
• 9. Sample-splitting, cross-fitting and repeated cross-fitting
  • 9.1. Cross-fitting with K folds
  • 9.2. Repeated cross-fitting with K folds and M repetitions
  • 9.3. Externally provide a sample splitting / partition
  • 9.4. Sample-splitting without cross-fitting
  • 9.5. Estimate DML models without sample-splitting
• 10. Sensitivity Analysis
  • 10.1. General algorithm
  • 10.2. Model-specific implementations