1 The Impact of Classification Errors on Assessing Software Inspection Processes with STAM

The Software Trouble Assessment Matrix (STAM) is a tool that software developers can use to evaluate the design and effectiveness of software inspection and testing processes so that they can be improved. STAM is used to organize relationships between thre e dimensions, Kenett (1994, 1995) and Kenett and Baker (1999). Three measures are computed from the data collected in a STAM analysis: • Negligence ratio: This ratio indicates the amount of errors that esca ped through the inspection process filters. In other words, it measures inspe ction efficiency. • Evaluation ratio: This ratio measures the delay of the inspection pro cess in identifying errors relative to the phase in which they occurred. In o ther words, it measures inspection effectiveness. • Prevention ratio: This ratio is an index of how early errors are dete cted in the development life cycle relative to the total number of reported erro rs. This is a combined measure of the development and inspection processes. It assesses the software developer's ability to generate and identify error s as early as possible in the development life cycle. Process improvements are the result of attempts too learnn f rom current andd past errors. Prerequisites too suchh improvement efforts are that the process es have beenn identifiedd andd that process ownershipp has beenn established.. A c ausal analysis of software faults classifies faults as being attributable too errors inn any one of these activities. Suchh efforts are sometimes labledd RCA (Root Cause Analysis) and ann exte nsive litterature onn the application of RCA inn a variety of industries is available (see refer ences).. For suchh ann analysis too be successful,, it is essential that there be agreement onn the boun daries of the processes represented by these activities.. Inn particular,, the entry andd exit criter ia for eachh process have too be clarified andd documented.. Suchh definitions permit effective data collection.. STAM is a methodd to analyze data derivedd by answering three questions: • Where were errors detectedd in the software life cycle ? • Where were those errors actually created? • Where couldd the errors have beenn detected? These three dimensions are positionedd like the letter T,, andd twoo check sheets are used too record the number of errors classifiedd inn the various combinations of the three dimensions. For example,, consider a certainn software versionn withh a total of 1100 reportedd errors at the completionn of acceptance …