Description: This project will build on a previously developed Bayesian ranking (BR) methodology [1, 2]. The BR was originally designed to support research funding agencies in their decision-making. However, it is broadly applicable to any context where limited resources have to be allocated under uncertainty, for example in prioritising drug candidates for further testing. In the context of funding decision-making, BR treats expert scores as noisy estimates of a proposal’s true quality and models them using a Bayesian hierarchical model. This allows us to rank proposals while explicitly accounting for the uncertainty in the process. When differences are small and uncertain, the approach might even suggest the use of a lottery. The BR is implemented in an R package using JAGS [3]. This thesis will extend the BR methodology to incorporate multiple sources of information. These could be preliminary and final scores, multiple evaluation criteria, primary and secondary decision criteria. The goal is to develop a Bayesian model that can summarise and synthesise this information into one single ranking. The proposed methodology will be implemented in R, possible in a new release of the R-package [3]. To illustrate and validate the methodology real-world data from funding decisions or drug trials will be used.

Contact: rachel.heyard@uzh.ch

References:
[1] Heyard, R., Ott, M., Salanti, G., & Egger, M. (2022). Rethinking the Funding Line at the Swiss National Science Foundation: Bayesian Ranking and Lottery. Statistics and Public Policy, 9(1), 110–121. https://doi.org/10.1080/2330443X.2022.2086190
[2] Heyard R, Pina DG, Buljan I, Marušić A (2025). Assessing the potential of a Bayesian ranking as an alternative to consensus meetings for decision making in research funding: A case study of Marie Skłodowska-Curie actions. PLoS ONE 20(3): e0317772. https://doi.org/10.1371/journal.pone.0317772
[3] Heyard R (2023). ERforResearch: Expected Rank for Research Evaluation. R package version 4.0.0. https://snsf-data.github.io/ERforResearch/