Abstract
This paper proposes a comprehensive suite of measures to quantify the level of class coupling during the design of object-oriented (OO) systems. This suite takes into account different OO design mechanisms, such as usage, specialization, and aggregation, thus capturing different kinds of coupling in OO systems. Based on data about operational failures of a commercial software system, our coupling measures are empirically investigated by analyzing their relationship with the probability of fault detection across classes. The results demonstrate that some of these coupling measures, along with some of Chi damber and Kemererüs measures, may be useful early quality indicators of the design of OO systems. In addition, principal component analysis shows that the underlying theory on which are based our coupling measures is partially supported by evidence. The results are then compared with a previous university study: we found a significant level of consistency and stability in the results.