Data

Tools

Indexes

Applications

Experiments

Open Source Science

https://github.com/bytedance/causalmatch

CausalMatch is a Bytedance research project aimed at integrating cutting-edge machine learning and econometrics methods to bring about automation in decision-making process.

https://github.com/bytedance/causalmatch

Science Score: 49.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • ○
    CITATION.cff file
  • ✓
    codemeta.json file
    Found codemeta.json file
  • ✓
    .zenodo.json file
    Found .zenodo.json file
  • ✓
    DOI references
    Found 1 DOI reference(s) in README
  • ✓
    Academic publication links
    Links to: arxiv.org
  • ○
    Committers with academic emails
  • ○
    Institutional organization owner
  • ○
    JOSS paper metadata
  • ○
    Scientific vocabulary similarity
    Low similarity (16.2%) to scientific vocabulary

Keywords

causal-inference econometrics machine-learning
Last synced: 7 months ago · JSON representation

Repository

CausalMatch is a Bytedance research project aimed at integrating cutting-edge machine learning and econometrics methods to bring about automation in decision-making process.

Basic Info
  • Host: GitHub
  • Owner: bytedance
  • License: apache-2.0
  • Language: Jupyter Notebook
  • Default Branch: main
  • Homepage:
  • Size: 754 KB
Statistics
  • Stars: 87
  • Watchers: 6
  • Forks: 5
  • Open Issues: 1
  • Releases: 5
Topics
causal-inference econometrics machine-learning
Created over 1 year ago · Last pushed 7 months ago
Metadata Files
Readme License

README.md

causamatch-logo CausalMatch: A Python Package for Propensity Score Matching and Coarsened Exact Matching

PyPI version Downloads Downloads Downloads

CausalMatch is a Python package that implements two classic matching methods, propensity score matching (PSM) and coarsened exact matching (CEM), to estimate average treatment effects from observational data. This package was designed and built as part of the ByteDance data science research program with the goal of combining state-of-the-art machine learning techniques with econometrics to bring automation to complex causal inference problems. Our toolkit possess the following features: * Implement classic matching techniques in the literature at the intersection of econometrics and machine learning * Maintain flexibility in modeling the propensity score model (via various machine learning classification models), while preserving the causal interpretation of the learned model and often offering valid confidence intervals * Use a unified API * Build on standard Python packages for Machine Learning and Data Analysis

Table of Contents - [News](#news) - [Getting Started](#getting-started) - [Installation](#installation) - [Usage Examples](#usage-examples) - [Estimation Methods](#estimation-methods) - [References](#references)

News

If you'd like to contribute to this project, please contact xiaoyuzhou@bytedance.com. If you have any questions, feel free to raise them in the issues section.

March 19, 2025: Release v0.0.5, see release notes here

Previous releases **December 10, 2024:** Release v0.0.4, see release notes [here](https://github.com/bytedance/CausalMatch/releases/tag/v0.0.4) **August 20, 2024:** Release v0.0.2, see release notes [here](https://github.com/bytedance/CausalMatch/releases/tag/v0.0.2) **August 2, 2024:** Release 0.0.1.

Getting Started

Installation

Install the latest release from [PyPI]: pip install causalmatch==0.0.5

Usage Examples

Estimation Methods

Propensity Score Matching (aka PSM) (click to expand) * Simple PSM ```Python from causalmatch import matching, gen_test_data from sklearn.ensemble import GradientBoostingClassifier df = gen_test_data(n = 10000, c_ratio=0.5) df.head() X = ['c_1', 'c_2', 'c_3', 'd_1', 'gender'] y = ['y', 'y2'] T = 'treatment' id = 'user_id' # STEP 1: initialize object match_obj = matching(data = df, T = T, X = X, y = y, id = id) # STEP 2: propensity score matching match_obj.psm(n_neighbors = 1, # number of neighbors model = GradientBoostingClassifier(), # p-score model trim_percentage = 0.1, # trim x percent of data based on propensity score caliper = 0.1) # caliper for p-score diff # STEP 3: balance check after propensity score matching match_obj.balance_check(include_discrete = True) # STEP 4: obtain average partial effect print(match_obj.ate()) ``` * PSM with multiple p-score models and select the best one based on f1 score ```Python # STEP 0: define all classification model you need from causalmatch import matching from sklearn.linear_model import LogisticRegression from sklearn.svm import SVC from sklearn.naive_bayes import GaussianNB from sklearn.neighbors import KNeighborsClassifier from sklearn.tree import DecisionTreeClassifier from sklearn.ensemble import RandomForestClassifier from sklearn.ensemble import GradientBoostingClassifier from lightgbm import LGBMClassifier from xgboost import XGBClassifier ps_model1 = LogisticRegression(C=1e6) ps_model2 = SVC(probability=True) ps_model3 = GaussianNB() ps_model4 = KNeighborsClassifier() ps_model5 = DecisionTreeClassifier() ps_model6 = RandomForestClassifier() ps_model7 = GradientBoostingClassifier() ps_model8 = LGBMClassifier() ps_model9 = XGBClassifier() model_list = [ps_model1, ps_model2, ps_model3, ps_model4, ps_model5, ps_model6, ps_model7, ps_model8, ps_model9] match_obj = matching(data = df, T = T, X = X, id = id) match_obj.psm(n_neighbors = 1, model_list = model_list, # input list of models you want to try trim_percentage = 0, caliper = 1, test_size = 0.2) # train-test split, what portion does test sample takes print(match_obj.balance_check(include_discrete = True)) df_out = match_obj.df_out_final_post_trim.merge(df[y + X + [id]], how='left', on = id) ```
Coarsened Exact Matching (click to expand) * Simple CEM ```Python match_obj_cem = matching(data = df, y = ['y'], T = 'treatment', X = ['c_1','d_1','d_3'], id = 'user_id') # coarsened exact matching match_obj_cem.cem(n_bins = 10, # number of bins for continuous x variables, cut by percentile k2k = True) # k2k: trim exp/base to have same observation numbers print(match_obj_cem.balance_check(include_discrete=True)) print(match_obj_cem.ate()) ``` * CEM with customized bin cut ```Python match_obj_cem = matching(data = df, y = ['y'], T = 'treatment', X = ['c_1','d_1','d_3'], id = 'user_id') match_obj_cem.cem(n_bins = 10, break_points = {'c_1': [-1, 0.3, 0.6, 2]}, # cut point for continuous variable cluster_criteria = {'d_1': [['apple','pear'],['cat','dog'],['bee']], 'd_3': [['0.0','1.0','2.0'], ['3.0','4.0','5.0'], ['6.0','7.0','8.0','9.0']]}, # group values for discrete variables k2k = True) ```

See the References section for more details.

References

S. Athey, J. Tibshirani, S. Wager. Generalized random forests. Annals of Statistics, 47, no. 2, 1148--1178, 2019.

V. Chernozhukov, D. Nekipelov, V. Semenova, V. Syrgkanis. Plug-in Regularized Estimation of High-Dimensional Parameters in Nonlinear Semiparametric Models. Arxiv preprint arxiv:1806.04823, 2018.

S. Wager, S. Athey. Estimation and Inference of Heterogeneous Treatment Effects using Random Forests. Journal of the American Statistical Association, 113:523, 1228-1242, 2018.

V. Chernozhukov, D. Chetverikov, M. Demirer, E. Duflo, C. Hansen, and a. W. Newey. Double Machine Learning for Treatment and Causal Parameters. ArXiv preprint arXiv:1608.00060, 2016.

Bajari, P., Burdick, B., Imbens, G. W., Masoero, L., McQueen, J., Richardson, T., & Rosen, I. M. (2021). Multiple randomization designs. arXiv preprint arXiv:2112.13495.

Owner

  • Name: Bytedance Inc.
  • Login: bytedance
  • Kind: organization
  • Location: Singapore

GitHub Events

Total
  • Release event: 3
  • Watch event: 47
  • Issue comment event: 1
  • Member event: 2
  • Push event: 20
  • Fork event: 4
  • Create event: 4
Last Year
  • Release event: 3
  • Watch event: 47
  • Issue comment event: 1
  • Member event: 2
  • Push event: 20
  • Fork event: 4
  • Create event: 4

Committers

Last synced: 11 months ago

All Time
  • Total Commits: 50
  • Total Committers: 4
  • Avg Commits per committer: 12.5
  • Development Distribution Score (DDS): 0.06
Past Year
  • Commits: 50
  • Committers: 4
  • Avg Commits per committer: 12.5
  • Development Distribution Score (DDS): 0.06
Top Committers
Name Email Commits
周小羽 x****u@b****m 47
lx-byte l****g@b****m 1
ajw-gelaoguan j****o@b****m 1
Willem Jiang 1****d 1
Committer Domains (Top 20 + Academic)
bytedance.com: 3

Issues and Pull Requests

Last synced: 8 months ago

All Time
  • Total issues: 1
  • Total pull requests: 1
  • Average time to close issues: N/A
  • Average time to close pull requests: less than a minute
  • Total issue authors: 1
  • Total pull request authors: 1
  • Average comments per issue: 0.0
  • Average comments per pull request: 0.0
  • Merged pull requests: 1
  • Bot issues: 0
  • Bot pull requests: 0
Past Year
  • Issues: 1
  • Pull requests: 1
  • Average time to close issues: N/A
  • Average time to close pull requests: less than a minute
  • Issue authors: 1
  • Pull request authors: 1
  • Average comments per issue: 0.0
  • Average comments per pull request: 0.0
  • Merged pull requests: 1
  • Bot issues: 0
  • Bot pull requests: 0
View more stats
Top Authors
Issue Authors
  • nickeubank (1)
Pull Request Authors
  • lx-byte (2)
Top Labels
Issue Labels
Pull Request Labels

Packages

  • Total packages: 1
  • Total downloads:
    • pypi 56 last-month
  • Total dependent packages: 0
  • Total dependent repositories: 0
  • Total versions: 6
  • Total maintainers: 2
pypi.org: causalmatch

Propensity score matching and coarsened exact matching

  • Versions: 6
  • Dependent Packages: 0
  • Dependent Repositories: 0
  • Downloads: 56 Last month
Rankings
Dependent packages count: 10.5%
Average: 34.9%
Dependent repos count: 59.3%
Maintainers (2)
langteam schouxy53
Last synced: 7 months ago

Dependencies

setup.py pypi