Departmental Research Assistant, ISERP Columbia
Applied causal-inference and count-data econometrics for policy research using quasi-Poisson regression, DiD design, and publication-ready empirical workflows.
Core Methods:Quasi-Poisson Regression, Difference-in-Differences, Robustness Diagnostics, Publication-Ready Visualization, Policy Evaluation Research Workflow: Specification Testing, Counterfactual Design, Empirical Validation, Reproducible Outputs Tooling: Python, Statistical Modelling, Data Visualization, Applied Social Science Inference
This work reflects an applied research role at the Institute for Social and Economic Research and Policy at Columbia University, where I supported empirical policy analysis under Professor Sandhya Kajeepeta. The emphasis was on producing causal evidence that could withstand specification scrutiny, rather than simply fitting models to data. The research environment demanded methodological discipline, reproducibility, and a clear translation of econometric findings into downstream academic deliverables.
A major component of the work involved applying quasi-Poisson regression to policy-relevant count outcomes, particularly in settings where variance structure made canonical Poisson assumptions too restrictive. I also used difference-in-differences designs to estimate treatment effects across competing empirical setups, testing whether conclusions remained stable under alternative identification choices and model specifications.
Beyond model estimation, I built robustness workflows that compared coefficients, inference stability, and fit across specifications, making it easier to distinguish durable empirical findings from fragile ones. This was important for ensuring that downstream conclusions were not artifacts of a single modelling choice, but reflected a genuinely defensible relationship in the data.
An equally important part of the role was communicating results. I produced publication-ready visualizations that converted technical output into interpretable research exhibits, helping bridge the gap between econometric analysis and the narrative demands of policy and academic writing. That meant treating presentation quality as part of the research process rather than as a cosmetic final step.
Overall, the role strengthened my ability to work at the intersection of causal inference, empirical rigor, and research communication. It also built a strong foundation for future work in policy evaluation, applied econometrics, and data-driven social science research.
Research Note: The exact project-level outputs were produced within an academic research setting, but the methods and responsibilities described here reflect the actual econometric and empirical workflow I supported.