Abstract | ||
---|---|---|
•Adversarial training may support causal inference from non-randomized data.•Minimization of the χ2-divergence between two populations minimizes the variance of importance sampling estimates.•The Counterfactual χ-GAN produces better feature balance, a less biased average treatment effect, and a more reasonable effective sample size than comparator methods.•When applied to real-world clinical data, The Counterfactual χ-GAN may provide an alternative means to causal estimation from observational data. |
Year | DOI | Venue |
---|---|---|
2020 | 10.1016/j.jbi.2020.103515 | Journal of Biomedical Informatics |
Keywords | DocType | Volume |
Causal inference,Machine learning,Health,GANs,Deep learning,Observational studies | Journal | 109 |
ISSN | Citations | PageRank |
1532-0464 | 0 | 0.34 |
References | Authors | |
0 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Amelia J Averitt | 1 | 0 | 0.34 |
Natnicha Vanitchanant | 2 | 0 | 0.34 |
Rajesh Ranganath | 3 | 0 | 2.37 |
Adler J. Perotte | 4 | 129 | 10.87 |