https://github.com/xuyiqing/fect_python

Python version of fixed effects counterfactual estimators
https://github.com/xuyiqing/fect_python

Last synced: 3 days ago
JSON representation

Python version of fixed effects counterfactual estimators

• Host: GitHub
• URL: https://github.com/xuyiqing/fect_python
• Owner: xuyiqing
• License: mit
• Created: 2022-11-04T08:39:58.000Z (about 2 years ago)
• Default Branch: main
• Last Pushed: 2024-07-25T09:26:22.000Z (4 months ago)
• Last Synced: 2024-08-02T06:02:27.918Z (3 months ago)
• Language: Jupyter Notebook
• Size: 5 MB
• Stars: 6
• Watchers: 1
• Forks: 1
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# fect_python

[![Lifecycle:

experimental](https://img.shields.io/badge/lifecycle-experimental-orange.svg)](https://www.tidyverse.org/lifecycle/#experimental)

[![License:

MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)

Fixed effects counterfactual estimator (**fect**) based on **rpy2**. This is a package for implementing counterfactual estimators in panel

fixed-effect settings. It is suitable for panel/TSCS analysis with

binary treatments under (hypothetically) baseline randomization. It

allows a treatment to switch on and off and limited carryover effects.

It supports linear factor models—hence, a generalization of

[**gsynth**](https://yiqingxu.org/packages/gsynth/index.html)—and the

matrix completion method.

**Repo:** [GitHub](https://github.com/xuyiqing/fect) (1.0.0)

**Examples:** The original R

[tutorial](https://yiqingxu.org/packages/fect/articles/tutorial.html) can be replicated in Python with [this](https://github.com/xuyiqing/fect_python/blob/main/example/fect_ipy.ipynb) tutorial. You can also find a markdown version tutorial [here](https://github.com/xuyiqing/fect_python/blob/main/fect_python_totorial_md/README.md)

**Reference:** Licheng Liu, Ye Wang, Yiqing Xu (2021). [A Practical

Guide to Counterfactual Estimators for Causal Inference with Time-Series

Cross-Sectional

Data](https://yiqingxu.org/papers/english/2022_fect/LWX2022.pdf).

*American Journal of Political Science*, conditionally accepted.

The original **R** package of **fect** can be found here: [GitHub](https://github.com/xuyiqing/fect) (1.0.0).

This package is based on [**rpy2**](https://rpy2.github.io/), which makes R objects available in Python environments.

## Installation

You can install **fect** from github by the following command:

```

pip install git+https://github.com/xuyiqing/fect_python.git

```

## Requirements

- Python 3.7+

- rpy2 3.5+

- numpy 1.1+

- pandas 1.1.2+

Although **fect_python** works on the latest version of **rpy2**, we strongly recommend installing **rpy2 3.5.4** in case of potential conflits.