Causal Evaluation and Learning of Treatment Rules under Partial Identification

PhD defence by Johannes Martin Hruza

Assessment Committee

[Professor Claus Thorn Ekstrøm, Department of Public Health, University of Copenhagen (Chairperson)
Professor Sören Möller, University of Southern Denmark
Professor Seth Flaxman, University of Oxford

Supervisors

Professor Samir Bhatt
Erin Evelyn Gabriel

Department

Department of Public Health

Graduate Programme

Biostatistics and Bioinformatics

Place

The defence is conducted as a hybrid defence.

To attend the defence in person:
Building: CSS 7, Room: CSS 7-0-34,
Øster Farimagsgade 5A, 1353 København

To attend the defence online:
Please follow the link to attend the defence online: https://ucph-ku.zoom.us/j/7909603643?pwd=ekNGRitXeDNyZFZYU2xIeS9qZklGQT09

Email address to gain access to the thesis: johannes.hruza@sund.ku.dk.
You will either receive a copy of the thesis or be informed where you can read a physical copy.
Recipients of copies of the thesis are not allowed to share or distribute it due to copyright compliance.

Short description of the thesis

Data-driven decision rules are used to guide choices in complex settings, from public policy to clinical practice. But a rule that predicts outcomes well does not necessarily tell us what would happen if a different action were taken. This thesis studies how to evaluate and learn individualized treatment rules: rules that assign actions based on individual characteristics. Such rules are often studied using observational data, where important factors may be unmeasured. This makes causal effects uncertain. Rather than hiding this uncertainty, the thesis develops methods that describe what can still be learned, contributing to more transparent and reliable decision-making from imperfect data.