Zachary Himmelsbach

I am a data scientist and statistician currently working in neuroscience and clinical psychology research. My other interests include developing methods for the analysis of unstructured data, causal inference, and educational research. I completed my AM and PhD at Harvard University.

Current Work

I develop methods for using natural language processing techniques and large language models in rigorous statistical analyses. My research has been funded by the Institute for Education Sciences and the Bill and Melinda Gates Foundation. Current projects include

  • Zero-shot learning methods coupled with generalized method of moments estimators to measure pedagogical practices from teachers’ speech data

  • High-dimensional word embeddings and permutation testing to identify words and phrases whose usage differs by group. I use this method to explore language differences between highly effective and less effective teachers

  • A computational method to assess the sensitivity of regression discontinuity results to possible data manipulation

Selected Publications

Bio

Previously, I worked as a research analyst at Harvard's Center for Education Policy Research under the mentorship of Doug Staiger and Thomas Kane. In that role, I developed statistical software for Bayesian analyses of rapid cycle randomized control trials. Since then, I have continued to develop statistical software; alongside Reagan Mozer and my PhD advisor Luke Miratrix, I contributed to RctText, an R package for impact analyses with text data as the outcome. My PhD commitee includes my advisor Sebastian Munoz-Najar, Luke Miratrix, and Andrew Ho.

I have also interned with J-PAL Global at MIT, worked at Conversica (a tech company), and served in the Peace Corps. My other collaborators include Dora Demzsky, Heather Hill, and Jing Liu.

If you're looking for a consultant or collaborator on research or analytic projects, feel free to get in touch. I’d love to hear from you.