Statistical Analysis of Two Experimental Studies

Software inspection is well-known as an eeective defect nding process. One avenue of inspection research concerns tool support, with the aim of further increasing its eeciency and eeectiveness. This has resulted in the production of a number of prototype tools. In general, however, these tools are poorly evaluated and the performance gain which these tools may provide has not been quantiied. This paper describes an incremental, cyclic approach to the development and evaluation of a tool to support software inspection. A version of the tool was developed then evaluated against traditional, paper-based inspection. The results were then used to guide development of the next version. This paper describes two of these cycles and the results of the corresponding evaluations. The approach was found to be a useful one, and is applicable to the development of many other tools, processes or procedures.