Measurement and Experimentation in Software Engineering.
Abstract:
The contributions of measurement and experimentation to the state-of-the-art in software engineering are reviewed. The role of measurement in developing theoretical models is discussed, and concerns for reliability and validity are stressed. Current approaches to measuring software characteristics are presented as examples. In particular, software complexity metrics related to control flow, module interconnectedness, and Halsteads Software Science are critiqued. The use of experimental methods in evaluating cause effect relationships is also discussed. Example programs of experimental research which investigated conditional statements and control flow are reviewed. The conclusion argues that advances in software engineering will be related to improvements in the measurement and experimental evaluation of software techniques and practices. Author