A comparative analysis of three replicated experiments comparing inspection and unit testing

Code inspections and structural unit testing are two common defect detection methods in software development. This paper analyses three replications of an experiment aimed at comparing which method finds more defects. The first two experiments use the same experimental design and materials, including the programs to be inspected or tested, while the third experiment replaces the programs with code for embedded systems. Participants are a mixture of graduate and undergraduate students. Our analysis shows that the both the differences in the instrumentation and the between-experiment participants themselves were larger than the differences between inspection versus unit testing, especially in terms of the time to complete the tasks.