microimpute

What problem does it solve?

Survey datasets often contain missing values, which can hinder accurate policy analysis. MicroImpute provides machine learning-based methods to intelligently fill these gaps while preserving statistical relationships, ensuring more complete and reliable data.

Core Features & Use Cases

• Multiple Imputation Methods: Supports linear, random forest, quantile forest, XGBoost, and hot deck methods to predict missing values.
• Quality Benchmarking: Compares different imputation methods using quantile loss to identify the most accurate approach for preserving data distribution.
• Hyperparameter Tuning: Integrates Optuna for automatic optimization of imputation model parameters.
• Use Case: Impute missing capital gains data in a survey dataset using a quantile forest model, leveraging patterns from a complete IRS tax record dataset, and then benchmark its quality.

Quick Start

Use the microimpute skill to impute missing capital_gains in your recipient DataFrame using donor data and income, age as common variables.