Counterfactual prediction from machine learning models: transportability and joint analysis for model development and evaluation using multi-source data
Both the observational and transportability analyses can be used to fit a model and estimate performance under a counterfactual treatment strategy in the population underlying the observational data, but they rely on different assumptions. In either case, the assumptions are untestable, and deciding which method is more appropriate requires careful contextual consideration. If all assumptions hold, then combining the data from the observational study and the randomized trial can be used for more efficient estimation.
Authors
- Christopher Boyer (ORCID: https://orcid.org/0000-0002-1101-2045)
- Issa J Dahabreh (ORCID: https://orcid.org/0000-0002-2215-9931)
- Habib Rahbar (ORCID: https://orcid.org/0000-0003-4835-1478)
- Despina Kontos (ORCID: https://orcid.org/0000-0001-9031-5126)
- Jon A. Steingrimsson (ORCID: https://orcid.org/0000-0003-2116-9377)
- Sarah Voter
Institutions
- Harvard University (US)
- University of Washington (US)
- Brown University (US)
- University of Pennsylvania (US)
Publication Details
- Journal
- Diagnostic and Prognostic Research
- Published
- 2025-10-01
- DOI
- https://doi.org/10.1186/s41512-025-00201-y
- Primary Topic
- Adversarial Robustness in Machine Learning
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Institutes of Health
- National Cancer Institute