A replicated quantitative analysis of fault distributions in complex software systems

To contribute to the body of empirical research on fault distributions during development of complex software systems, a replication of a study of Fenton and Ohlsson is conducted. The hypotheses from the original study are investigated using data taken from an environment that differs in terms of system size, project duration, and programming language. We have investigated four sets of hypotheses on data from three successive telecommunications projects: 1) the Pareto principle, that is, a small number of modules contain a majority of the faults ( in the replication, the Pareto principle is confirmed), 2) fault persistence between test phases ( a high fault incidence in function testing is shown to imply the same in system testing, as well as prerelease versus postrelease fault incidence), 3) the relation between number of faults and lines of code ( the size relation from the original study could be neither confirmed nor disproved in the replication), and 4) fault density similarities across test phases and projects ( in the replication study, fault densities are confirmed to be similar across projects). Through this replication study, we have contributed to what is known on fault distributions, which seem to be stable across environments.