Statistical Methods for Data Cleaning with the BeEST Phase-III Heavy Neutrino Search

The aim of the Beryllium Electron capture in Superconducting Tunnel junction (BeEST) experiment is to identify theorized heavy neutrino mass eigenstates. This is investigated via the electron capture (EC) decay of ⁷Be by measuring the recoil energy spectrum of the ⁷Li product. To enable precise modeling of the decay's spectral characteristics, large amounts of data are needed. In the BeEST Phase-III, data was collected via hours-long measurements taken in many separate sessions. Consequently, slight differences in experimental conditions may be propagated into the datasets themselves. In this work, we present on statistical methods used to identify and correct (where possible) disagreements in spectral calibration and energy resolution for these datasets; this is done primarily by conducting statistical tests on K-shell peaks of the EC spectra measured in different experiments. Intra-experiment testing is also undertaken to search for temporally dependent degradation effects. The results allow us to restore the viability of some miscalibrated datasets and provide a statistical basis for which sets to exclude from model fitting.