Code for the article "netivreg: Estimation of Peer Effects in Endogenous Social Networks"

The command netivreg implements the Generalized Three-Stage Least Squares (G3SLS) estimator developed in Estrada (2022), 'Causal Inference in Multilayered Networks,' Ph.D. Dissertation and the Generalized Method of Moments (GMM) estimator in Chan, T. J., J. Estrada, K. P. Huynh, D. T. Jacho-Chávez, C. T. Lam, and L. Sánchez-Aragón (2024) 'On the Estimation of Social Effects with Observational Network Data and Random Assignment,' Journal of Econometric Methods for the endogenous linear-in-means model. The two procedures utilize full observability of a two-layered multiplex network data structure using Stata's new multiframes capabilities and Python integration (version 16 or above).

Authors:

Pablo Estrada
Emory University, Atlanta, GA, USA

Juan Estrada
Analysis Group Economic Consulting, Washington, DC, USA

Kim Huynh
Bank of Canada, Ottawa, Ontario, Canada

David Jacho-Chávez
Emory University, Atlanta, GA, USA

Leonardo Sánchez-Aragón
ESPOL University, Guayaquil, Guayas, Ecuador

Requirements

Stata 16.0 or above, and Python 3.9 or above are needed as well as the following Python packages,

Python Package	Version
networkX	3.2.x
numpy	1.26.x
pandas	2.2.x
scikit-learn	1.5.x
scipy	1.13.x

Step-by-Step Installation Guide for netivreg

1. Install Python via Anaconda:

   Download and install Anaconda on your computer, which comes with Python 3.9 or higher. The Anaconda distribution is highly recommended to ensure compatibility with netivreg.

   Note: On Windows, Anaconda is typically installed in the C:\Users\<username>\anaconda3 folder.

2. Create a Python environment for netivreg:

   To ensure backward compatibility and avoid version conflicts, we recommend you create a dedicated environment for netivreg with the required Python packages. Do this by first, openning the Anaconda prompt and run the following command:

   conda create -n netivreg_env python=3.9 pip networkx=3.2 numpy=1.26 pandas=2.2 scikit-learn=1.5 scipy=1.13

   This command creates a conda environment named netivreg_env with Python 3.9 and the necessary packages with the compatible versions.

   Note: On Windows, this environment is typically created within the C:\Users\<username>\anaconda3\envs folder. Inside envs, a subfolder named netivreg_env will hold all files for this specific environment.

3. Activate the Python environment for netivreg:

   Run the following command on the Anaconda prompt:

   conda activate netivreg_env

4. Directing Stata to use the crated Python environment:

   After completing steps 1–3 above, link Python to your local Stata installation. Open Stata and use the following command in the Stata command window to set the Python version:

   Note: On Windows, one would write

   python set exec "C:\Users\<username>\anaconda3\envs\netivreg_env\python.exe" , permanently

   Replace <username> with your actual username in the path. This command tells Stata to use the specified Python executable within the Anaconda environment netivreg_env. Setting it permanently means Stata will remember this configuration for future sessions. Otherwise, you will need to set this address every time you would like to use the netivreg command.

Manual Installation:

5. Copy netivreg files to the Stata ado folder:

   • Locate the netivreg.ado and netivreg.sthlp files in the Files folder within this repository.
   • Copy these files to the /ado/plus/n folder in your local user installation directory. For example, on Windows, this path would be C:/Users/<username>/ado/plus/n. This folder is automatically created when you install any user-written Stata package.

6. Copy Python Files to the Stata ado/plus/py Folder:

   • Locate the ado/plus/py folder in your local user installation directory. For example, C:/Users/<username>/ado/plus/py in Windows. If this folder does not exist, you can create it.
   • Inside the ado/plus/py folder, create a new folder named netivreg.
   • Locate the __init__.py, g3sls.py and gmm.py files in the Files folder within this repository.
   • Copy these files to the ado/plus/py/netivreg folder.

   Optional Setup: Directly Adding Python Files to ado/plus/py (We thank Stata for making this suggestion.)

   • Change the content of __init__.py to the following:

      __version__ = "1.0.0"
      from g3sls import *
      from gmm import *

   • Previously, we used from .g3sls import * and from .gmm import *, with a . prefix. This update removes the . to simplify the import statements.

   • Rename the new __init__.py to netivreg.py.

   • Copy the netivreg.py, g3sls.py, and gmm.py into the folder ado/plus/py. Now, you do not need to create the folder named netivreg.

How-To Video

For Windows users, please refer to the following instructional video

Tips

• To recover the location path of the environment netivreg_env, the user can run the following command in the command prompt,

  conda env list

• To desactive an environment, in your Aanconda Prompt run

  conda desactivate

Usage

In Files folder within this repository, there is a do-file named manuscript_output.do that replicates the tables from the manuscript 'netivreg: Estimation of Peer Effects in Endogenous Social Networks' using the datasets found in the same Files folder in this repository.