A Conceptual Basis for Software Engineering Measurement

The ten publications chosen for inclusion in this PhD by Publication submission have resulted from work on (or following on from) the ESPRIT II project “METKIT”, which began in early 1989 and ended in mid-1992. The aim of METKIT was to produce a range of teaching materials about software metrics, or - more accurately - software engineering measurement (a new term adopted by METKIT), for use within industry and academia. The project began with an in-depth study of current practice in both industry and academia, as well as a literature survey, which revealed the need for some form of conceptual framework to pull together the apparently diverse range of topics making up the subject of software metrics. Existing frameworks were found to be inadequate due to ambiguity and lack of rigour. In particular, it was felt that there was a need to relate these diverse topics to measurement theory, which was increasingly seen as playing a central role. Also, the boundary of the subject area itself was found to be ill-defined. The subsequent definition and wide-spread dissemination (both in industry and academia) of a coherent software engineering measurement culture has been the primary achievement of METKIT. The publications contained in this submission indicate the strong impact which the METKIT teaching materials have had (up to the time of writing), as well as presenting the details of the original conceptual framework which forms their basis. The development of this conceptual framework came about through careful consideration of the definition of measurement, and also adoption of goal-driven measurement as a central notion. This led on to the identification of a natural boundary for the subject of software engineering measurement. For example, the consideration of a broad range of measurement goals resulted in the inclusion of topics such as experimental design, how to go about setting up a measurement programme and human issues in data collection. Previously, the subject of software metrics was seen as being centred on ways of measuring software, with an emphasis on complexity measurement. This is now seen more as a niche area in software engineering measurement.