A 3+1 Sterile Neutrino Global Fit using Simulation-Based Inference

The neutrino sector has long been subject to unexplained anomalies that may be resolved through one or more sterile neutrinos. Sterile neutrinos can oscillate to active neutrino flavors, disrupting expected 3-flavor oscillation probabilities. The addition of one sterile neutrino to the Standard Model is called a 3+1 model, which leads to a 5σ improvement in fits to data. However, there is large tension between neutrino appearance and disappearance datasets, suggesting there is no preferred set of parameters that fits the data well. While past global fits have relied on Wilks' theorem, the oscillatory nature of the signal violates the assumptions required for its validity. The Feldman-Cousins approach is more accurate, though requires computing a test-statistic for many pseudoexperiments which is computationally infeasible in a global fit. In this work I present 3+1 global fit results to short-baseline neutrino data using simulation-based inference to generate an approximated test-statistic. I demonstrate the frequentist validity of this method and show that it achieves orders-of-magnitude improvements in computing speed. I discuss fits to appearance and disappearance channels, evaluate tension, and explore individual experiments' results within a 3+1 framework. This analysis is an updated assessment of how well the 3+1 model accounts for long-standing neutrino anomalies without reliance on Wilks' theorem.