DeepBlip: Estimating Conditional Average Treatment Effects Over Time

Haorui Ma, Dennis Frauen, Stefan Feuerriegel · 2025

Read DeepBlip: Estimating Conditional Average Treatment Effects Over Time by Haorui Ma, Dennis Frauen, Stefan Feuerriegel (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-14545.txt.

Description

Haorui Ma, Dennis Frauen, and Stefan Feuerriegel present DeepBlip, a neural method to estimate treatment effects that unfold over time. Structural nested mean models split a full treatment sequence effect into local blip effects at each time. That split aids interpretation and fast policy evaluation. Past versions could not be trained end to end with neural networks.

A double optimization trick lets all blip functions learn at once. Sequence networks like LSTMs or transformers capture complex time patterns. The method adjusts for time varying confounding by design. Its orthogonal loss adds robustness when nuisance models are imperfect. Experiments test the approach on time based treatment data.

Catalog record

CREATORHaorui Ma, Dennis Frauen, Stefan Feuerriegel
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE135 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-14545

Browse the full library · About the Tower of Babel