diddesign

Difference-in-differences (DID) remains the workhorse design for observational causal inference in the social sciences. When multiple pre-treatment periods are available, the standard two-period DID and sequential DID estimators exploit different moment conditions, each valid under its own parallel-trends assumption. diddesign implements the Egami and Yamauchi (2023, Political Analysis) GMM framework that combines these component estimators with efficient weighting, yielding the Double DID estimator.

The package provides a unified Python interface for standard DID, sequential DID, Double DID, K-DID (three or more pre-treatment periods with adaptive J-test moment selection), and staggered-adoption designs with lead-specific estimates. All results are returned as immutable objects with pandas DataFrame accessors for inspection, export, LaTeX table generation, and figure production.

Analysts working with panel data or repeated cross-sections will find a complete workflow: pre-treatment placebo diagnostics via equivalence confidence intervals, treatment effect estimation with bootstrap inference, GMM weight inspection, and publication-ready output in a single package.

Getting started

Install with pip install diddesign, then run the Quick Start example in User Guide using the built-in Malesky (2014) dataset. Full workflows are available in Tutorials; concise snippets in Examples.

Contents:

• User Guide
  • Installation
  • Quick Start
  • Data Requirements
  • Designs
  • Parameter Selection Guide
  • Result Objects
  • Reading Returned Rows
  • Interpreting Results
  • Formula Interface
  • Covariates
  • Inference
  • Visualization
  • Built-in Datasets
  • Error Handling
  • Bootstrap Inference for Staggered Adoption
  • Summary Statistic Edge Cases
  • Plotting Row Requirements
  • Formula Validation
  • Time Label Validation
  • Boolean Numeric Validation
  • Random Seed Validation
  • Design Role and Cluster Validation
  • Bootstrap and Sequence Validation
  • GMM Availability Consistency
  • Negative GMM Weights
  • Boolean as_frame Validation
  • Estimator Filter Validation
  • Sequential DID Transformed Outcomes
  • SA Component Timing Weights
  • SA Placebo Timing Weights
  • Dataset Registry
• Tutorials
  • Minimal RCS Double DID
  • Inspecting Component Weights
  • Real-Data RCS Workflow (Malesky 2014)
  • Staggered Adoption Workflow (Paglayan 2019)
  • K-DID with J-test
  • Preparing Plotting Rows
  • LaTeX Export
  • Using Covariates
• Examples
  • Built-in Datasets
  • RCS Double DID (Malesky 2014)
  • Staggered Adoption (Paglayan 2019)
  • K-DID with J-test (Paglayan 2019)
  • RCS with Covariates
  • Formula Interface
  • Visualization
  • LaTeX Table Export
  • Running Repository Examples
  • Dataset Provenance
• API Reference
  • Core Estimation
  • Result Objects
  • Summary and Reporting
  • Formula Parsing
  • Plotting Rows
  • Visualization
  • Data Loading
  • Errors and Validation
  • Package Metadata

Indices and tables

• Index

• Module Index

• Search Page