Ds Synthetic Control

Synthetic Control Methods

Comprehensive guide to synthetic control methods in machine learning and data science workflows.

When to Use This Skill

• Solving real-world causal inference problems
• Building machine learning pipelines with synthetic control methods
• Implementing best practices for synthetic control methods
• Optimizing model performance using synthetic control methods techniques
• Learning industry-standard approaches to synthetic control methods

When NOT to Use This Skill

• When using pre-built libraries without understanding underlying concepts
• For toy problems that don't require synthetic control methods rigor
• When domain expertise in specific problem requires different approach
• If your problem doesn't require the complexity this skill provides

Purpose and Key Concepts

Synthetic Control Methods is a critical component of the machine learning workflow. This skill covers:

1. Theoretical foundations — Mathematical principles and statistical concepts
2. Practical implementation — Working code examples and patterns
3. Common pitfalls — Mistakes to avoid and how to recover from them
4. Best practices — Industry-standard approaches and optimization techniques

Core Workflow

1. Understand the problem — Clearly define what you're solving for
2. Select approach — Choose the right technique for your data and constraints
3. Implement solution — Write clean, tested code following best practices
4. Validate results — Verify your implementation with tests and validation
5. Optimize performance — Improve efficiency and accuracy incrementally

Implementation Patterns

Pattern 1: Basic Synthetic Control Methods

import numpy as np
from scipy.optimize import minimize
from typing import Tuple

def compute_synthetic_weights(
    X_donor: np.ndarray, 
    X_treated: np.ndarray, 
    regularization: float = 1e-4
) -> np.ndarray:
    """
    Compute optimal donor weights for synthetic control using constrained optimization.
    
    Args:
        X_donor: (n_donors, n_features) matrix of donor pre-treatment data
        X_treated: (n_features,) vector of treated unit pre-treatment data
        regularization: L2 penalty to prevent overfitting weights
        
    Returns:
        Optimal non-negative weights summing to 1.0
        
    Raises:
        ValueError: If input dimensions are incompatible
    """
    if X_donor.shape[1] != X_treated.shape[0]:
        raise ValueError("Donor features must match treated feature dimensions")
        
    def objective(w: np.ndarray) -> float:
        prediction = X_donor @ w
        residual = X_treated - prediction
        return float(np.sum(residual**2) + regularization * np.sum(w**2))
        
    constraints = (
        {'type': 'eq', 'fun': lambda w: np.sum(w) - 1.0}
        {'type': 'ineq', 'fun': lambda w: w}
    )
    
    n_donors = X_donor.shape[0]
    initial_weights = np.ones(n_donors) / n_donors
    
    result = minimize(
        objective, 
        initial_weights, 
        constraints=constraints, 
        method='SLSQP'
        options={'maxiter': 1000, 'ftol': 1e-9}
    )
    
    if not result.success:
        raise RuntimeError(f"Weight optimization failed: {result.message}")
        
    return result.x

Pattern 2: Production-Ready Synthetic Control Methods

import logging
import numpy as np
import pandas as pd
from typing import Dict, Any, Optional
from scipy.optimize import minimize

logger = logging.getLogger(__name__)

class SyntheticControlEstimator:
    """Production-grade Synthetic Control Method estimator with DiD support."""
    
    def __init__(self, regularization: float = 1e-4, max_iter: int = 1000) -> None:
        self.regularization = regularization
        self.max_iter = max_iter
        self.weights_: Optional[np.ndarray] = None
        self.pre_error_: Optional[float] = None
        self.post_error_: Optional[float] = None
        
    def _optimize_weights(self, X_donor: np.ndarray, X_treated: np.ndarray) -> np.ndarray:
        """Find non-negative weights summing to 1 that minimize pre-treatment error."""
        def loss(w: np.ndarray) -> float:
            pred = X_donor @ w
            error = X_treated - pred
            return float(np.sum(error**2) + self.regularization * np.sum(w**2))
            
        constraints = (
            {'type': 'eq', 'fun': lambda w: np.sum(w) - 1.0}
            {'type': 'ineq', 'fun': lambda w: w}
        )
        
        n_donors = X_donor.shape[0]
        w0 = np.ones(n_donors) / n_donors
        
        res = minimize(
            loss, w0, constraints=constraints, method='SLSQP'
            options={'maxiter': self.max_iter}
        )
        if not res.success:
            logger.warning(f"Weight optimization did not converge: {res.message}")
        return res.x
        
    def fit_predict(
        self, 
        donor_data: pd.DataFrame, 
        treated_data: pd.DataFrame, 
        period_col: str = 'period', 
        treated_unit_id: str = 'treated'
    ) -> Dict[str, Any]:
        """Fit SCM and predict counterfactual outcomes."""
        if donor_data.empty or treated_data.empty:
            raise ValueError("Input DataFrames cannot be empty")
            
        donor_pre = donor_data[donor_data[period_col] < 0].drop(columns=[treated_unit_id])
        treated_pre = treated_data[treated_data[period_col] < 0].drop(columns=[treated_unit_id])
        
        X_donor = donor_pre.values
        X_treated = treated_pre.values.flatten()
        
        self.weights_ = self._optimize_weights(X_donor, X_treated)
        self.pre_error_ = float(np.mean((X_treated - X_donor @ self.weights_)**2))
        
        donor_post = donor_data[donor_data[period_col] >= 0].drop(columns=[treated_unit_id])
        counterfactual = donor_post.values @ self.weights_
        
        return {
            'weights': self.weights_
            'pre_mse': self.pre_error_
            'counterfactual': counterfactual
            'status': 'success'
        }

Best Practices

• ✅ Always validate your implementation on test data
• ✅ Document your assumptions and methodology
• ✅ Use version control for reproducibility
• ✅ Monitor performance metrics in production
• ✅ Periodically review and update your approach
• ✅ Test with edge cases and outliers
• ✅ Log all significant operations for debugging

# BAD: Hardcoded values, no validation, violates DRY principle
def bad_scm(data):
    w = [0.1, 0.2, 0.3, 0.4]  # Magic numbers
    return sum(w) * data['y']

# GOOD: Type hints, validation, configurable parameters, follows SOLID/DRY
def good_scm(
    donor_matrix: np.ndarray, 
    target_vector: np.ndarray, 
    reg_lambda: float = 1e-3
) -> np.ndarray:
    if donor_matrix.shape[1] != target_vector.shape[0]:
        raise ValueError("Dimension mismatch")
    weights = compute_synthetic_weights(donor_matrix, target_vector, reg_lambda)
    return weights

Common Pitfalls

| Pitfall | Problem | Solution | |

Constraints

MUST DO

• Validate all data preprocessing steps are fit-only on training data, never on validation or test sets
• Implement reproducible pipelines with fixed random seeds and deterministic operations where possible
• Report model performance with confidence intervals via bootstrapping or cross-validation across multiple runs
• Log all experiments with parameters, metrics, and artifacts using MLflow or equivalent tracking system

MUST NOT DO

• Do not evaluate a model on the same data used for training — always hold out a proper test set
• Avoid overfitting to the validation set by limiting hyperparameter search iterations
• Never use features that can only be computed at inference time (look-ahead bias)
• Do not report single-run accuracy without statistical significance testing or error bars

Live References

Authoritative documentation links for this skill's domain. The model follows markdown links at load time to resolve external references and inline content.

• Synthetic Control Method — Wikipedia
• CausalImpact — Bayesian Structural Time Series (Google)
• Synthetic Controls for Policy Evaluation (NBER Paper)
• Double Machine Learning for Causal Inference
• Time Series Counterfactuals — Microsoft DoWhy