Demonstrating the usage of single-case designs in experimental software engineering

Abstract

Experimental software engineering is the subdiscipline of empirical software engineering which uses experimentation to analyze, improve, and to validate software engineering methods (concepts, techniques, models). The efficacy and efficiency of some methods have already been demonstrated by experiments. To achieve this factorial designs were used so far. Critics of software engineering experiments argue that doing an experiment would be incredibly expensive. Furthermore, for doing software engineering experiments one would need hundreds of subjects. This paper shows that this is not necessarily right. An approach is presented that shows how to conduct software engineering experiments in a cost-effective way. The approach is useful to analyze software engineering problems specific to one subject, to conduct pilot experiments that precede in-depth experiments, and to accompany technology transfer. To demonstrate the usage of a single-subject experiment the domain of reuse was chosen.