microcalibrate

What problem does it solve?

Survey data often needs adjustment to accurately represent population totals (benchmarks), and traditional calibration can result in dense, inefficient datasets. MicroCalibrate addresses this by providing advanced calibration with sparsity and hyperparameter tuning.

Core Features & Use Cases

• Weighted Survey Calibration: Adjusts survey weights to precisely match known population targets (e.g., total income, employment).
• Sparsity with L0 Regularization: Encourages many weights to become zero, reducing dataset size for faster simulations without losing accuracy.
• Automatic Hyperparameter Tuning: Optimizes calibration parameters to achieve the best balance between target matching and sparsity.
• Use Case: Calibrate a survey dataset's household weights to match known population totals for income and employment, reducing the number of active households by 60% for faster PolicyEngine simulations.

Quick Start

Use the microcalibrate skill to calibrate your weights to match targets using your estimate_matrix with a sparsity penalty.