An EmpincalStudy of Software Metrics
H. F. Li, William K. Cheung · 1987
metrics arecomputed forthepurpose ofevaluating certain characteristics ofthesoftware developed. AFortran static source codeanalyzer, FORTRANAL,wasdeveloped tostudy 31metrics, including anewhybrid metric introduced inthis paper, andapplied toadatabase of255programs, allofwhichwerestudent assignments. Comparisons amongthese metrics areperformed. Their crosscorrelation confirms theinternal consistency ofsomeofthese metrics whichbelong tothesameclass. Toremedytheincompleteness ofmost ofthese metrics, theproposed metric incorporates context sensitivity tostructural attributes extracted fromaflowgraph. Itisalsoconciuded that'many volumemetrics havesimilar performance while some control metrics surprisingly correlate wellwithtypical volumemetrics inthetest samples used.A flexible class ofhybrid metric canincorporate bothvolumeandcontrol attributes inassessing software complexity. IndexTerms-Control complexity, crosscorrelation, empirical study, Halstead's software science, hybrid metrics, software metric.