Causal Inference

Time-series causal inference and statistical validation methods.

Applied Causal Inference

released

Dual-language causal methods codebase with tests and validation architecture. Implements DoubleML, instrumental variables, and time-series-aware variants in both Julia and Python. The intent is reference-quality implementations with explicit validation against synthetic data and known answers.

Stack: Julia · Python · DoubleML · CausalImpact

What's next

Development paused at reference quality. Future direction: add additional estimators and integrate with the temporal-validation utilities.

double_ml_time_series

in progress

Implementation and methodology notes for double machine learning extended to time-series settings. Covers temporal partially-linear DML, cross-fitting protocols that respect time ordering, and HAC-style inference. Synthetic examples illustrate where naive cross-fitting fails and what to do instead.

Stack: Python · DoubleML · statsmodels · LaTeX

Part of the Books & Guides corpus →

What's next

All 10 chapters of the web edition are live (potential outcomes + Frisch–Waugh–Lovell through the research pipeline); the manuscript also ships as a PDF in the repo. Next: revision passes as the methodology evolves.

TemporalValidation / temporalcv

released

Utilities for temporally-aware cross-validation: walk-forward splits with explicit gaps, purging adjacent observations, embargo windows, and other leakage-prevention patterns. The methodology lineage notes connect classical temporal CV practice to modern uses in causal and ML pipelines.

Stack: Julia · Python

What's next

Published to PyPI as temporalcv. Future: a Julia companion module with a consistent API.