Speaker
Description
Hadronically decaying tau leptons are reconstructed as narrow jets, and quark or gluon-initiated jets that are misidentified as hadronic taus are a dominant background in searches with hadronic tau final states. Since the misidentification probability depends on the least well-modelled aspects of the simulation (soft fragmentation) and this background can't be taken from Monte Carlo and must be measured in data. The idea of this talk is to give an overview of the data-driven methods used in ATLAS for this purpose (matrix method, fake factors and the universal fake factor method), and then focus on the new, continuous fake factor method and its application within the Type II (and III) seesaw analysis. A common limitation of the standard approaches is that they are binned in a few variables and assume that the control region reproduces the fake composition of the signal region, which is usually the dominant systematic uncertainty. Instead the continuous method estimates the fake factor as a event density ratio (in event) learned from data in a high-dimensional feature space.