/boostIV

Causal Gradient Boosting: Boosted Instrumental Variable Regression

Primary LanguagePythonOtherNOASSERTION

Intro

Code for

Causal Gradient Boosting: Boosted Instrumental Variable Regression
Edvard Bakhitov and Amandeep Singh
https://arxiv.org/abs/2101.06078

Installation

Before running any script you need to install the boostIV package by running

python setup.py install

The main dependencies are numpy, sklearn and numdifftools which will be automatically downloaded by running the setup.py command above.

boostIV package

This will install the package boostIV on your local python environment, which contains boostIV and post-boostIV estimators from the paper.

Third party material

The folder DeepGMM contains the source files from the git repo https://github.com/CausalML/DeepGMM that contains the code from the work of Bennet et al. "Deep Generalized Method of Moments for Instrumental Variable Analysis".

The code in monte_carlo/KIV.py is based on the Matlab codes from the git repo https://github.com/r4hu1-5in9h/KIV for the paper by Singh et al. "Kernel Instrumental Variable Regression".