Research

Publications


I am looking forward to filling this section very soon.


Work in Progress


Inference for Conditional Average Treatment Effects using Distributional Nearest Neighbors

– Ongoing Research Project
– GitHub: Repository containing Working Paper and R-Package
– Presented at: New York Camp Econometrics XIX (Poster)

I consider the problem of Inference for the Distributional / Layered Nearest Neighbor Estimator for nonparametric regression and extend its applicability to the estimation of Conditional Average Treatment Effects. Using recently developed theory for generalized U-Statistics, I develop results to allow for less restrictive assumptions to justify inference in the regression setting and develop asymptotic theory in the presence of functional nuisance parameters using a doubly robust approach and cross fitting. This effectively results in a novel way to implement the second stage of a doubly robust estimator nonparametrically.


Variance Estimation for Hájek-Dominated Generalized U-Statistics

– Ongoing Research Project

I show that under a single, checkable condition, Hájek Projection Dominance, the classic jackknife is ratio-consistent for variance estimation across Generalized U-Statistics, including randomized, incomplete, and infinite-order settings. This reduces the problem to verifying that the first-order projection governs variance. The framework connects classical resampling to modern ensemble learners and clarifies when jackknife variance estimates are reliable.
I verify the conditions for the two-scale distributional nearest-neighbor estimator, yielding practical confidence intervals and improving on previously available results.

Distribution-on-Distribution Regression with Optimal Transport

– Ongoing Research Project

We study Regression Methods in settings with Distribution-Valued Random Variables. With this goal in mind we employ methods from Optimal Transport, a subfield of mathematics that has gained a lot of traction in recent years. The core idea is to embrace the geometry of the underlying Wasserstein-Space when performing regression. This idea allows us to obtain more intuitive and arguably better results than usual linear regression using tools from functional data analysis alone. In Economics this allows us to better understand how multiple distributions, such as income and wealth distributions, are related.


TBA

– Joint work with: Harold D. Chiang and Christian Döbler

I am looking forward to sharing more details about this project in the future.


Previous Projects


For Projects from my Master’s and Bachelor’s Program look here.