← Toutes les offres
Inria

Internship: Development of non-parametric methods for the evaluation of surrogate markers in the non-continuous data setting

Inria

Bordeaux
Stage
Internet And Software
Publié le 17 novembre 2026

Description du poste

*Le descriptif de l’offre ci-dessous est en Anglais* **Type de contrat :** Convention de stage **Niveau de diplôme exigé :** Bac + 5 ou équivalent **Fonction :** Stagiaire de la recherche **Contexte et atouts du poste** ------------------------------- The main objective of a clinical trial is to evaluate the effect of a treatment on a clinical endpoint. However, direct measurement of the clinical endpoint may not be feasible due to cost, time, or ethical reasons. In such cases, one may instead rely on a surrogate marker, a variable which can be used to predict treatment effects on the clinical endpoint. Evaluating candidate surrogate markers is a long-standing and active area of statistical research. Conventional approaches rely on parametric models whose distributional assumptions are often hard to verify, and inference can be misleading when they fail. This is a real concern for modern biomedical data, which are frequently complex, noisy, and high-dimensional. To address this, some authors have recently proposed non-parametric methods using rank- based tests for surrogate evaluation (Parast et al., 2024). These methods do not make as- sumptions about the underlying form of the data, provide interpretable and scale-free surrogacy metrics, and are valid in small samples. This makes them particularly well-suited to high-dimensional applications, such as transcriptomic datasets, which are becoming more commonplace in modern biomedical studies (Hughes et al., 2025, 2026). A limitation of these approaches is that they have only been developed for the evaluation of continuous variables, both for the clinical endpoint and the surrogate candidates. However, endpoints are often of different types, such as binary/ordinal data (e.g. protection status after vaccination), or time-to-event data (e.g. time to cancer progression). Extending the framework to these settings is not straightforward, since ties and censoring complicate rank- based inference, but it would substantially broaden the range of applications. Send a detailed CV and a motivation letter to Boris Hejblum (boris.hejblum@u-bordeaux.fr) and Arthur Hughes (arthur.hughes@u-bordeaux.fr). **Mission confiée** ------------------- **References** Arthur Hughes, Layla Parast, Rodolphe Thiébaut, and Boris P. Hejblum. Rise: Two‐stage rank‐based identification of high‐dimensional surrogate markers applied to vaccinology. Statistics in Medicine, 44(20-22), September 2025. ISSN 1097-0258. doi:10.1002/sim.70241. Arthur Hughes, Rodolphe Thiébaut, Layla Parast, and Boris P. Hejblum. Meta- analysis of high-dimensional surrogate markers. arXiv preprint, 2026. doi:https://doi.org/10.48550/arXiv.2605.03819. Gemma Moncunill, Jason Carnes, William Chad Young, Lindsay Carpp, Stephen De Rosa, Joseph J Campo, Augusto Nhabomba, Maxmillian Mpina, Chenjerai Jairoce, Greg Finak, Paige Haas, Carl Muriel, Phu Van, Héctor Sanz, Sheetij Dutta, Benjamin Mordmüller, Selidji T Agnandji, Núria Díez-Padrisa