Nonparametric Assessment of the Calibration of Individualized Treatment Effects
Mohsen Sadatsafavi, Jeroen Hoogland, Thomas P. A. Debray, John Petkau · 2026
Read Nonparametric Assessment of the Calibration of Individualized Treatment Effects by Mohsen Sadatsafavi, Jeroen Hoogland, Thomas P. A. Debray, John Petkau (2026) 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/pmc-13504708.txt.
Description
Mohsen Sadatsafavi, Jeroen Hoogland, Thomas P A Debray, and John Petkau build the first proper test of whether a personalized treatment model is honest about its own predictions. Their 2026 paper defines the target: a model is moderately calibrated if, among patients it predicts will gain a given amount from treatment, the true average gain equals that amount. Two obstacles have blocked testing it: each patient's untreated outcome is never seen, and the predictions are continuous.
Their solution tracks cumulative prediction errors as a process whose long-run behavior mirrors Brownian motion, letting standard limit theory supply the tests. Two constructions are offered, one leaning on the model's risk predictions and one purely marginal. Simulations confirm the method catches miscalibration of several shapes. A case study shows how to present the results, and an R package ships with the paper. The authors aim the tool at anyone validating individualized medicine models.
Catalog record
| CREATOR | Mohsen Sadatsafavi, Jeroen Hoogland, Thomas P. A. Debray, John Petkau |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 64 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |