py-why/dowhy

Features

• Causal Inference Frameworks - Structures causal analysis into a sequential four-step framework of modeling, identification, estimation, and refutation.
• Causal Inference Tools - An open-source Python library for estimating causal effects from observational data using a unified four-step framework.
• Causal Counterfactual Answerers - Computes what would have happened under alternative scenarios by intervening on the causal model and simulating outcomes.
• Causal Identification-Estimation Separators - Decouples the causal identification problem from statistical estimation, allowing any compatible estimator for a given estimand.
• Causal Effect Estimators - Estimates the magnitude of a causal relationship from observational data using a unified interface that separates identification from estimation.
• Causal Estimand Identifiers - Determines the target estimand for a causal effect based on the given graph and data.
• Causal Estimate Refuters - Validates the robustness of a causal estimate through automated checks such as adding a random common cause.
• Refutation Testing Tools - Includes automated refutation testing that validates causal estimates by introducing random common causes or placebo treatments.
• Causal Effect Estimation Interfaces - Estimates causal effects by separating identification from estimation, supporting methods like double machine learning.
• Causal Counterfactual Simulators - Provides a counterfactual simulation engine that computes alternative outcomes by intervening on causal models.
• Causal Assumption Validators - Runs sensitivity and robustness checks on estimates, falsifies causal graphs, and evaluates fitted causal mechanisms to verify assumptions.
• Causal Identification-Estimation Separators - Distinguishes the causal problem from the statistical problem, allowing any estimator to be used for a target estimand.
• Causal Effect Estimators - Provides a unified interface for estimating causal effect magnitudes from observational data using identification-estimation separation.
• Causal Assumption Modelers - Explicitly models causal assumptions as first-class citizens, viewable graphically or as conditional independence statements, and tests them against data.
• Causal Graph Creators - Creates a causal model from data and a user-defined graph, specifying treatment and outcome variables.
• Directed Acyclic Graph Models - Represents causal assumptions as directed acyclic graphs that encode variable relationships and support graphical validation.
• Graph Validation Tools - Ships tools for creating and validating causal graphs from data with automatic assumption testing against observations.
• Causal Graph Validations - Automatically tests whether the specified causal graph matches observed conditional independence patterns in the data.
• Causal Assumption Validators - Automatically tests whether stated causal assumptions hold by running sensitivity and robustness checks on the obtained estimate.
• Causal Estimate Refutations - Ships automated refutation tests that systematically introduce violations to validate the robustness of causal estimates.
• Causal Algorithm Applicators - Selects appropriate identification methods and causal mechanisms by default while allowing full customization of parameters and models.
• Heterogeneous Treatment Effect Estimators - Implements double machine learning and other methods for estimating heterogeneous treatment effects across subpopulations.
• Causal Model Interventional Simulators - Provides a counterfactual simulation engine that computes alternative outcomes by intervening on causal models.
• Causal Machine Learning Integrations - Integrates double machine learning with orthogonal moment conditions for debiased heterogeneous treatment effect estimation.
• Causal Structure Learners - Learns and diagnoses causal relationships among variables, including root cause analysis and modeling of data generation processes.
• Structure Learning Algorithms - Ships algorithms for learning causal structures from data, including root cause analysis and data generation process modeling.
• Causal Inference - Estimate causal effects.
• Statistical Modeling - Library for causal inference and testing causal assumptions.

Historial de estrellas