I will discuss three developments addressing three fundamental issues found in the statistical analysis of particle physics data. The first has to do with combining or interpreting results of experiments, where the likelihood of the data given some physically meaningful parameters is the key object. I will describe the recent progress in the 20-year quest to publish these likelihoods. The second topic involves the reinterpretation of results that cannot be addressed by the likelihood alone as it requires processing alternative signal hypotheses through the analysis pipeline. I will describe the recent progress in RECAST, which was proposed 10 years ago. Finally, I will describe a challenge at the heart of analyzing HEP data and reformulate traditional HEP analysis in terms of likelihood-free inference (or simulation-based inference, as I prefer to call it). I will describe how machine learning is pushing the frontier of simulation-based inference and how it can greatly enhance the sensitivity of measurements at the LHC (eg. for constraining effective field theories).
The seminar will be done remote only.
M. Girone, M. Elsing, L. Moneta, M. Pierini
Event co-organised with the PHYSTAT Committee