Visualizing Event Studies

Here you’ll find a visualization tool designed for researchers interested in analyzing event studies. This tool allows users to define their estimands, impose their assumptions, and select valid observations for analysis in their datasets. Additionally, the platform includes a suite of weighting diagnostics, describing the implied weighted populations, group-wise effective sample size (ESS), information contributions of observation groups, and influence of individual observations on the estimator. This visualization tool aims to enhance the transparency of event study analyses and empower researchers to make well-informed decisions in their analysis.

---

Table of Contents

1. Introduction
2. Features
3. Installation
4. Contributing
5. License

---

Introduction

This platform is an interactive visualization tool for researchers working on event studies. It enables users to:

• For a given dataset of event studies with staggered intervention, implement the classical dynamic TWFE regression estimator and a proposed robust weighting estimator developed in our recent paper.

• Perform diagnostics to visualize and understand the implied target population, analyze effective sample size (ESS) and information borrowing.

The platform is based on R and Shiny to offer real-time interaction, allowing users to dynamically adjust parameters and observe results.

---

Features

1. Customizable Parameters: import your dataset (e.g., CSV files) for analysis, specify in the dataset the variable name corresponding to the outcome, the binary treatment, the unit of analysis (e.g., country, state, county), and the time (e.g., year, quarter, month). Note: we will not keep a copy of your dataset. All the interactive analysis can be conducted locally.

2. Interactive Visualizations: we provide the following visualizations to guide you through the analysis of event studies.

3. Real-Time Data Analysis and Diagnostics:

4. Downloadable Results: You will have the option to download the diagnostic measures.

---

Installation

To run this software locally, ensure that you have R and the required packages installed.

Prerequisites

• R (version >= 4.0)

• RStudio

• Shiny package: Install using install.packages("shiny")

• Required R packages for analysis including:

Steps

1. Clone this repository:

git clone https://github.com/zhushen3128/EventStudy.git
cd EventStudy

2. Open app.R in RStudio. Remember to save the util.R file in the same directory.

3. Run the app using:

library(shiny)
runApp('app.R')

Usage

• Start the Application: Run the Shiny app as described above.

• Upload Data: Import datasets (e.g., CSV files) for analysis. In the GitHub repository, we provide a sample dataset named bacon_cleaned.cvs which is used as the default dataset for the tool.

• Explore Features: Use the interactive controls to adjust assumptions, modify parameters, and visualize outcomes.

• Export Results: Save plots and analysis outputs for reporting.

---

Contributing

We welcome contributions to improve this platform! To contribute:

• Fork this repository.

• Create a new branch:

git checkout -b feature-name

• Commit your changes and submit a pull request.

---

License

This project is licensed under the MIT License.