PanelSFA: High-Performance Stochastic Frontier Analysis in Python

PanelSFA is a modern Python library designed to fill the “Python Gap” in econometric software. It brings production-grade Stochastic Frontier Analysis (SFA) to the scientific computing ecosystem with a focus on numerical stability, scikit-learn compatibility, and high-performance panel data estimation.

By replicating the capabilities of proprietary software like Stata, PanelSFA democratizes technical efficiency analysis for researchers and practitioners alike.

Key Features

• Comprehensive Model Suite: Implements foundational models like ALS (1977) and BC95, alongside advanced panel estimators like Greene’s True Fixed Effects and the Four-Component model.

• Scikit-learn API: Designed with an Estimator API for seamless integration into machine learning pipelines and hyperparameter searches.

• Numerical Stability: Utilizes OLS warm-starts, parameter reparameterization (log/logit links), and L-BFGS-B optimization for reliable convergence on complex likelihood surfaces.

• Performance: Optimized for large panel datasets using vectorized operations and efficient memory management.

Getting Started

• Installation
  • Install via PyPI
• Quickstart Guide
  • Cross-Sectional Data (ALS 1977)
  • Panel Data (True Fixed Effects)

Models & Methodology

• Supported Models
  • Cross-Sectional Models
  • Panel Models (Deterministic)
  • Panel Models (Unobserved Heterogeneity)
  • The Four-Component Model
• Theory & Methodology
  • The Error Decomposition
  • Numerical Stability Engineering

API Reference

• API Reference
  • Base Estimator
  • Cross-Sectional Models
  • Standard Panel Models
  • Unobserved Heterogeneity (Greene 2005)
  • Multi-Component Models

Indices and tables

• Index

• Module Index

• Search Page