Statistical Analysis on Software Metrics Affecting Modularity in Open Source Software

Modularity has been identified by many researchers as one of the success factors of Open Source Software (OSS) Projects. This modularity trait are influenced by some aspects of software metrics such as size, complexity, cohesion, and coupling. In this research, we analyze the software metrics such as Size Metrics (NCLOC, Lines, and Statements), Complexity Metrics (McCabe's Cyclomatic Complexity), Cohesion Metrics (LCOM4), and Coupling Metrics (RFC, Afferent coupling and Efferent coupling) of 59 Java-based OSS Projects from Sourceforge.net. By assuming that the number of downloads can be used as the indication of success of these projects, the OSS Projects being selected are the projects which have been downloaded more than 100,000 times. The software metrics reflecting the modularity of these projects are collected using SONAR tool and then statistically analyzed using scatter graph, Pearson r product-moment correlation, and least-square-fit linear approximation. It can be shown that there are only three independent metrics reflecting modularity which are NCLOC, LCOM4, and Afferent Coupling, whereas there is also one inconclusive result regarding Efferent Couplin