Effectiveness and Efficiency of Fault Removal Techniques

由于缺乏控制和复制阻碍了许多软件工程研究，因此需要设计利用已完成项目（以及相关数据）的实验，而不是设计“实时”项目（将产品推出的目标可能与数据捕获和分析目标相竞争）。该研究项目包括构思、设计和准备一系列大规模、多地点的回顾性实验。研究集中在四种故障识别技术（需求检查、正式规范、正式设计和故障注入）上，通过确定哪些方法在什么环境下更有效，从而建立在现有工作的基础上。特别是，该项目寻求证据，证明哪种方法在故障识别和排除方面最有效和/或最有效。这项研究涉及一个由工业界和学术界合作伙伴组成的联盟，这些合作伙伴都隶属于霍华德大学软件技术评估研究中心。在可能的情况下，将通过与学生项目一起进行部分实验来测试部分实验设计，以确认其可行性并帮助指导准备工作。结果应该有广泛的影响，无论是对实践者（他们会发现结果对指导资源分配决策有用）和研究人员（他们会注意到用回顾性数据进行软件工程领域实验的可行性）。

Because lack of control and replication impede many software engineering studies, there is a need to design experiments that take advantage of completed projects (and the associated data), rather than "live" projects (where the goal of getting the product out the door may compete with data capture and analysis goals). The research project involves conceiving, designing and preparing a collection of large-scale, multi-location retrospective experiments. The research focuses on four fault identification techniques (requirements inspection, formal specification, formal design, and fault injection), building upon existing work by determining which methods are more effective and in what contexts. In particular, the project seeks evidence about which method is most effective and/or efficient at fault identification and removal. The research involves a consortium of industrial and academic partners, all affiliated with Howard University's Center for Research in Evaluating Software Technology. Where possible, parts of the experimental design will be tested by running parts of the experiment with student projects to confirm its feasibility and to help direct the preparatory efforts. The results should have wide impact, both on the community of practitioners (who will find the results useful for guiding resource allocation decisions) and researchers (who will note the feasibility of doing software engineering field experiments with retrospective data).