Skip to content

quickstart.rst

Latest commit

 

History

History
208 lines (146 loc) · 4.74 KB

quickstart.rst

File metadata and controls

208 lines (146 loc) · 4.74 KB

Getting Started

This guide will help you get started with diff-diff for Difference-in-Differences analysis.

Installation

Install diff-diff using pip:

pip install diff-diff

Basic 2x2 DiD

The simplest DiD design has two groups (treated/control) and two periods (pre/post).

import pandas as pd
from diff_diff import DifferenceInDifferences, generate_did_data

Tip

Most estimators have short aliases for convenience — e.g. from diff_diff import DiD, TWFE, CS, DDD. See the :doc:`API reference <api/estimators>` for the full list.

# Generate synthetic data with a known treatment effect
data = generate_did_data(
    n_units=100,
    n_periods=10,
    treatment_effect=5.0,
    treatment_period=5,
    treatment_fraction=0.5,
)

# Fit the model
did = DifferenceInDifferences()
results = did.fit(
    data,
    outcome='outcome',
    treatment='treated',
    time='post'
)

# View results
print(results.summary())

Output:

Difference-in-Differences Results
==================================
ATT:           5.123
Std. Error:    0.456
t-statistic:   11.23
p-value:       0.000
95% CI:        [4.229, 6.017]

Using Formula Interface

You can also use R-style formulas:

did = DifferenceInDifferences()
results = did.fit(data, formula='outcome ~ treated * post')

Adding Covariates

Control for confounders with the covariates parameter:

results = did.fit(
    data,
    outcome='outcome',
    treatment='treated',
    time='post',
    covariates=['age', 'income']
)

Cluster-Robust Standard Errors

For panel data, cluster standard errors at the unit level:

did = DifferenceInDifferences(cluster='unit_id')
results = did.fit(data, outcome='y', treatment='treated', time='post')

Two-Way Fixed Effects

For panel data with multiple periods:

from diff_diff import TwoWayFixedEffects

twfe = TwoWayFixedEffects()
results = twfe.fit(
    data,
    outcome='outcome',
    treatment='treated',
    unit='unit_id',
    time='period'
)

Event Study Design

Examine treatment effects over time:

from diff_diff import MultiPeriodDiD

event = MultiPeriodDiD()
results = event.fit(
    data,
    outcome='outcome',
    treatment='treated',
    time='period',
    post_periods=[5, 6, 7, 8, 9],
    reference_period=4
)

# Plot the event study
from diff_diff.visualization import plot_event_study
ax = plot_event_study(results)

Staggered Adoption

When treatment is adopted at different times across units:

from diff_diff import CallawaySantAnna

cs = CallawaySantAnna()
results = cs.fit(
    data,
    outcome='outcome',
    unit='unit_id',
    time='period',
    first_treat='first_treat'
)

# View aggregated treatment effect
print(f"Overall ATT: {results.overall_att:.3f}")

Parallel Trends Testing

Test the key identifying assumption:

from diff_diff.utils import check_parallel_trends

trends_result = check_parallel_trends(
    data,
    outcome='outcome',
    time='period',
    treatment_group='treated',
    pre_periods=[0, 1, 2, 3]
)

if trends_result['p_value'] > 0.05:
    print("Parallel trends assumption supported")

Sensitivity Analysis

Assess robustness to parallel trends violations with Honest DiD:

from diff_diff import HonestDiD

# Compute bounds under relative magnitudes restriction
honest = HonestDiD(method="relative_magnitude", M=1.0)
bounds = honest.fit(event_study_results)

print(f"Robust CI: [{bounds.ci_lb:.3f}, {bounds.ci_ub:.3f}]")

Next Steps