Using only treated units and an unlabeled pool, positive-unlabeled learning can select control groups whose ATE estimates match the true effect in simulations and two agricultural case studies.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Positive-Unlabeled Learning for Control Group Construction in Observational Causal Inference
Using only treated units and an unlabeled pool, positive-unlabeled learning can select control groups whose ATE estimates match the true effect in simulations and two agricultural case studies.