rdd

rdd is a set of tools for implementing regression discontinuity designs in Python. At present, it only allows for inputs that are pandas Series or DataFrames. Check out the tutorial here for a guide to using this package.

Installation:

To install using pip, run pip install rdd. See here for more information and to download from PyPI. Otherwise, you can download this repository.

Current Features:

• Computes the Imbens-Kalyanaraman optimal bandwidth (see this article for details)
• Creates a dataset containing only observations within a given bandwidth
• Generates a reduced pandas DataFrame for graphical analysis (such as binned scatter plots) when the number of observations is large
• Implements a sharp RDD
  • With or without user supplied equation and/or controls
  • Via OLS or WLS

Features to Come:

• Tutorial on how to check rdd's statistical assumptions in Python, such as using:
  • Continuity plots
  • Density plots
  • Tests for discontinuities
  • Checking for balance and testing for random assignment (though not a requirement, it can still be a useful check)
• McCrary tests
• Implementation of Fuzzy RDD