A statistical estimation of the coupling between object metric for open-source apps developed in Java

Abstract

The coupling between objects along with other metrics, is used for evaluating the faults, vulnerabilities, and other quality indicators in software systems, including open-source ones. It is known, that a coupling between objects value between one and four is good. However, there are apps in Java for which the coupling between objects metric value at an app level is greater than four. That is why, in our opinion, the above interval for coupling between objects needs to be clarified for the app level. To find the recommended values for the coupling between objects mean of an app we have proposed to apply the confidence and prediction intervals. A coupling between objects mean value of an app from the confidence interval is good since this interval indicates how reliable the estimate is for all apps. A coupling between objects mean value higher than an upper bound of the prediction interval may indicate that some classes are too tightly coupled with other ones in the app. We have estimated the confidence and prediction intervals of the coupling between objects mean using normalizing transformations for the data sample from one hundred open-source apps developed in Java hosted on GitHub. Comparison with the coupling between objects mean values of three popular open-source apps developed in Java illustrate the applicability of the proposed quality indicators in the form of the confidence and prediction intervals of the coupling between objects mean.