Abstract
An analysis of the existing data on the atmospheric neutrino anomaly is presented, focused on the statistical significance that can be attributed to its experimental evidence. Our approach is alternative to the usual analyses in terms of the μ/e ratio of event rates. In fact, we perform a comparison between data and expectations, by separating the information on e-like and μ-like events, with a careful estimate of the different errors and of their correlation effects. The results are shown both numerically and graphically, and disclose interesting aspects of the atmospheric neutrino anomaly that the use of the μ/e ratio would partially hide, both in the sub-GeV and in the multi-GeV energy range.