A predictive model for identifying inspections with escaped defects

Defect origin and inspection cost data pertaining to a large software development project was analyzed by using statistical techniques. First, we analyzed the difference between the inspections with escaped defects and those without escapes regarding inspection hours, preparation hours, and code size. Next, we examined the composite measures such as, total effect, preparation/inspection ratio, and effort ratio. Then, we developed a linear discriminant function to assess the effectiveness of an inspection. The results lead to several findings regarding the effect of total inspection effort, preparation time, and artifact size on the occurrence of defect escapes. The methodology used in this study can be applied to other software development organizations, and the findings can be generalized to a variety of inspection processes