Large mass ice/water Cherenkov experiments, optimized to detect low energy (1-20 GeV) atmospheric neutrinos, have the potential to discriminate between normal and inverted neutrino mass hierarchies. The sensitivity depends on several model and detector parameters, such as the neutrino flux profile and normalization, the Earth density profile, the oscillation parameter uncertainties, and the detector effective mass and resolution.A proper evaluation of the mass hierarchy discrimination power requires a robust statistical approach. In this work, the Toy Monte Carlo, based on an extended unbinned likelihood ratio test statistic, was used. The effect of each model and detector parameter, as well as the required detector exposure, was then studied. While uncertainties on the Earth density and atmospheric neutrino flux profiles were found to have a minor impact on the mass hierarchy discrimination, the flux normalization, as well as some of the oscillation parameter (∆m 2 31 , θ 13 , θ 23 , and δ CP ) uncertainties and correlations resulted critical. Finally, the minimum required detector exposure, the optimization of the low energy threshold, and the detector resolutions were also investigated.