Combining Data from Reading Experiments in Software Inspections: A Feasibility Study

Software inspections have been around for 25 years, and most software engineering researchers and professionals know that they are mostly a cost-effective means for removing software defects. However, this does not mean that there is consensus about how they should be conducted in terms of reading techniques, number of reviewers or the effectiveness of reviewers. Still, software inspections are probably the most extensively empirically studied technique in software engineering. Thus, a large body of knowledge is available in literature. This paper uses 30 data sets from software inspections found in the literature to study different aspects of software inspections. As a feasibility study, the data are amalgamated to increase our understanding and illustrate what could be achieved if we manage to conduct studies where a combination of data can be collected. It is shown how the combinated data may help to evaluate the influence of several different aspects, including reading techniques, team sizes and professionals vs. students. The objective is primarily to illustrate how more general knowledge may be gained by combining data from several studies. It is concluded that combining data is possible, although there are potential validity threats. Research results are examined with reference to software inspections on three levels: organization, project and individual.