Software Quality Prediction Models Compared

Numerous empirical studies confirm that many software metrics aggregated in software quality prediction models are valid predictors for qualities of general interest like maintainability and correctness. Even these general quality models differ quite a bit, which raises the question: Do the differences matter? The goal of our study is to answer this question for a selection of quality models that have previously been published in empirical studies. We compare these quality models statistically by applying them to the same set of software systems, i.e., to altogether 328 versions of 11 open-source software systems. Finally, we draw conclusions from quality assessment using the different quality models, i.e., we calculate a quality trend and compare these conclusions statistically. We identify significant differences among the quality models. Hence, the selection of the quality model has influence on the quality assessment of software based on software metrics.