| Code based measurements are frequently used by practitioners during software development. This study examines code based measurements for object oriented systems from the perspective of a software development manager. Based on past research a theoretical model relating code based Object Oriented system design metrics and product quality was developed. Hypotheses based on the model were tested using a data sample of 55 releases from 12 different Open Source projects. The study identified four key findings. First, code structure decides the strength of the relationship between metrics and defect count. Second, aggregation effect converts weak relationships at the class level into strong relationships at the package level. Third, when adequate testing is done there is a very good chance of finding correlation at the package level between metrics and defect count. Fourth, at the package level, number of lines of code is as good a measure as other object oriented metrics. Based on these key findings the study has recommended steps for using these metrics in practice. |
