Deficit hawks: robust new physics searches with unknown backgrounds

Abstract

Searches for new physics often face unknown backgrounds, causing false detections or weakened upper limits. This paper introduces the deficit hawk technique, which mitigates unknown backgrounds by testing multiple options for data cuts, such as fiducial volumes or energy thresholds. Combining the power of likelihood ratios with the robustness of the interval-searching techniques, deficit hawks could improve mean upper limits on new physics by a factor two for experiments with partial or speculative background knowledge. Deficit hawks are well-suited to analyses that use machine learning or other multidimensional discrimination techniques, and can be extended to permit discoveries in regions without unknown background.