CausalPy - causal inference for quasi-experiments
A Python package focussing on causal inference for quasi-experiments. The package allows users to use different model types. Sophisticated Bayesian methods can be used, harnessing the power of PyMC and ArviZ. But users can also use more traditional Ordinary Least Squares estimation methods via scikit-learn models.
Installation
To get the latest release:
pip install CausalPy
Alternatively, if you want the very latest version of the package you can install from GitHub:
pip install git+https://github.com/pymc-labs/CausalPy.git
Quickstart
import causalpy as cp
# Import and process data
df = (
cp.load_data("drinking")
.rename(columns={"agecell": "age"})
.assign(treated=lambda df_: df_.age > 21)
)
# Run the analysis
result = cp.pymc_experiments.RegressionDiscontinuity(
df,
formula="all ~ 1 + age + treated",
running_variable_name="age",
model=cp.pymc_models.LinearRegression(),
treatment_threshold=21,
)
# Visualize outputs
fig, ax = result.plot();
# Get a results summary
result.summary()
Videos
Features
Different quasi-experimental methods
Rather than focussing on one particular quasi-experimental setting, this package aims to have broad applicability. We can analyse data from the following quasi-experimental methods:
Synthetic control
This is appropriate when you have multiple units, one of which is treated. You build a synthetic control as a weighted combination of the untreated units.
Geographical Lift / Geolift
We can also use synthetic control methods to analyse data from geographical lift studies. For example, we can try to evaluate the causal impact of an intervention (e.g. a marketing campaign) run in one geographical area by using control geographical areas which are similar to the intervention area but which did not recieve the specific marketing intervention.
ANCOVA
This is appropriate when you have a single pre and post intervention measurement and have a treament and a control group.
Difference in differences
This is appropriate when you have pre and post intervention measurement(s) and have a treament and a control group.
Regression discontinuity
Regression discontinuity designs are used when treatment is applied to units according to a cutoff on a running variable, which is typically not time. By looking for the presence of a discontinuity at the precise point of the treatment cutoff then we can make causal claims about the potential impact of the treatment.
Support
This repository is supported by PyMC Labs.
