Extracting relative thresholds for source code metrics

Abstract

Meaningful thresholds are needed for promoting software metrics as an effective instrument to measure the internal quality of systems. To address this challenge, we propose in this thesis the concept of relative thresholds for evaluating metrics data following heavy-tailed distributions. The proposed concept assumes that metric thresholds should be followed by most entities, but that it is also natural to have entities in the 'long-tail' that do not follow the defined limits. We describe an empirical method for deriving relative thresholds from a set of systems and apply this method to the well-known Qualitas Corpus. We compare the proposed thresholds with thresholds extracted according to a method largely used by the software industry. We also evaluate our method under different contexts and using the history of versions of five systems. Finally, we validate our method with software practitioners. Overall, we found that good quality systems as cited by experts follow the thresholds.