Econometrust
High-performance econometric regression library written in Rust with Python bindings.
Overview
Econometrust provides optimized implementations of fundamental econometric estimators for applied research and production applications. The library combines computational efficiency with statistical rigor, offering comprehensive inference capabilities and robust numerical algorithms.
Core Estimators
- OLS - Ordinary Least Squares with robust standard errors
- Ridge - Ridge Regression with L2 regularization for multicollinearity
- WLS - Weighted Least Squares for heteroskedastic models
- GLS - Generalized Least Squares for correlated errors
- IV - Instrumental Variables for endogeneity
- TSLS - Two-Stage Least Squares for overidentified models
- FE - Fixed Effects for panel data
Key Features
Statistical Capabilities - Comprehensive statistical inference (standard errors, t-tests, confidence intervals) - Multiple R-squared measures including adjusted R-squared - Professional summary output with diagnostic statistics - Robust and cluster-robust standard error options
Computational Performance - Rust-based backend for maximum speed - Memory-optimized algorithms with intelligent selection - Vectorized operations using optimized linear algebra - Numerically stable implementations with fallback methods
API Design
- Consistent interface across all estimators
- Comprehensive type annotations and documentation
- Error handling with informative diagnostics
- Integration with NumPy ecosystem
Installation
Basic Usage
import numpy as np
import econometrust
# Generate data
X = np.random.randn(100, 3)
y = X @ [1.5, -2.0, 0.5] + np.random.randn(100) * 0.1
# Fit model
model = econometrust.OLS(fit_intercept=True, robust=True)
model.fit(X, y)
# Results
print(f"R-squared: {model.r_squared:.4f}")
print(f"Adjusted R-squared: {model.r_squared_adj:.4f}")
print(model.summary())
Documentation
Getting Started
- Installation - Setup and requirements
- Quickstart - Basic tutorial
API Reference
- Overview - Method comparison and selection guide
- Individual estimator documentation with mathematical foundations
Estimator Selection Guide
| Data Structure | Recommended Method | Key Requirement |
|---|---|---|
| Cross-sectional, no heteroskedasticity | OLS | Homoskedastic errors |
| Multicollinearity or overfitting concerns | Ridge | Need for regularization |
| Known heteroskedasticity | WLS | Specified weights |
| Serial/spatial correlation | GLS | Known covariance structure |
| Endogenous regressors (exact ID) | IV | Valid instruments |
| Endogenous regressors (over ID) | TSLS | Multiple instruments |
| Panel data | FE | Entity-time structure |