ISERN: A Distributed Experiment - Ein verteiltes Inspektionsexperiment

Abstract

In the long run, high quality of software is prerequisite for software companies to survive. Inspections of software products help to detect and remove errors in software development early and cost-effectively. Thus, they help to enhance the quality of software products. One main goal in practice is, for a given situation, to select a suitable inspection technique. Empirical studies are necessary to make such a selection as they provide real-world data from a documented context.

The results of particular empirical studies, however, are not easily transferable to other situations without knowledge about their situation, their context. Thereby, it is especially unclear to which degree experimental results depend on their context, that is, which context factors influence the results. Insight into experimental results in different contexts can, for example, be gained by repeating, or replicating, experiments in these contexts. Thus, distributed experiments that are conducted in many different contexts and that systematically measure context factors are especially desirable, as they can improve the generalization of experimental results.

This paper describes a concept for a distributed inspection experiment being planned and conducted by the International Software Engineering Research Network (ISERN). Thereby, about 20 organizations from industry and research collaborate. The first step of the empirical study is a broad survey of software companies to reveal the state of the practice in the areas of inspection process and inspection techniques. The second step is to conduct experiments with documents from within ISERN and state-of-the-art inspection techniques, as well as with the participating organization’s own documents and inspection techniques. Goal of the distributed experiment is to describe where the variations in practice of inspections techniques are and to observe the effectiveness of these inspection techniques with respect to relevant context factors. From a research perspective, this experiment can help find out how experimental results can be generalized more effectively.