API Reference Overview

Econometrust provides six main estimator classes, each designed for specific econometric modeling scenarios.

Estimator Classes

Linear Regression Models

Class	Description	Use Case
OLS	Ordinary Least Squares	Standard linear regression with optional robust standard errors
Ridge	Ridge Regression (L2 Regularization)	Regularized regression for multicollinearity and overfitting prevention
WLS	Weighted Least Squares	Heteroskedastic models with known variance weights
GLS	Generalized Least Squares	Models with known error covariance structure

Instrumental Variables Models

Class	Description	Use Case
IV	Instrumental Variables	Exactly identified models with endogenous regressors
TSLS	Two-Stage Least Squares	Overidentified models with multiple instruments

Panel Data Models

Class	Description	Use Case
FE	Fixed Effects	Panel data with unobserved entity heterogeneity

Common Interface

All estimators share a consistent interface:

Initialization

model = EstimatorClass(fit_intercept=True, **kwargs)

Fitting

model.fit(X, y, **additional_args)

Properties and Methods

• model.coefficients - Regression coefficients
• model.intercept - Intercept term (if fitted)
• model.predict(X) - Generate predictions
• model.standard_errors() - Coefficient standard errors
• model.t_statistics() - T-statistics for significance tests
• model.p_values() - P-values for hypothesis tests
• model.confidence_intervals(alpha) - Confidence intervals
• model.summary() - Comprehensive regression output
• model.r_squared - Coefficient of determination
• model.residuals - Regression residuals
• model.mse - Mean squared error

Algorithm Selection

Econometrust automatically selects the most appropriate numerical algorithm based on:

• Problem size (number of observations and features)
• Matrix conditioning (numerical stability requirements)
• Available computational resources

Available Algorithms

1. Cholesky Decomposition: For well-conditioned overdetermined systems
2. SVD (Singular Value Decomposition): For ill-conditioned or small systems
3. Normal Equations: For large overdetermined systems with good conditioning

This intelligent selection ensures optimal performance and numerical stability across different problem types.