Public portfolio of graduate-level projects in causal inference, Bayesian modeling, multilevel analysis, and agent-based simulation. The work in this repository is implemented primarily in Python and Jupyter notebooks, with an emphasis on interpretable methods, policy-relevant questions, and clear written communication.
Built an agent-based Monte Carlo simulation on weighted social networks to study the feedback loop between social isolation, recommendation systems, ideological drift, and friendship decay. The project compares interventions that reduce harmful algorithmic amplification and strengthen corrective social support.
- Notebook: CS166_final_codebook.ipynb
- Report: CS166_final_report_with_figures.pdf
- Simulation preview: simulation.gif
Applies experimental and observational causal inference methods to estimate treatment effects. The notebook covers randomized evaluation, matching, instrumental variables, and synthetic control, including an analysis of job-training outcomes and a synthetic-control study of the California drought's effect on crime.
- Notebook: Causal Inference.ipynb
Uses probabilistic modeling to compare student test-score distributions across years when the observed data are available only as histograms with inconsistent bins. The project emphasizes partial pooling, model comparison, and uncertainty-aware inference.
- Notebook: Discrete and multi-level models.ipynb
Develops Bayesian regression models for property pricing in Buenos Aires, comparing Normal, Student-t, and outlier-aware specifications to improve robustness under heavy tails and unusual observations.
- Notebook: bayesian_regression_real_estate_BA_code.ipynb
- Report: Bayesian Regression for Real Estate Pricing in Buenos Aires (PDF)
- Bayesian regression and probabilistic modeling
- Causal inference and policy evaluation
- Hierarchical and discrete outcome models
- Agent-based simulation on social networks