Evaluating the accuracy of defect estimation models based on inspection data from two inspection cycles

  • Authors:

  • Stefan Biffl;Wilfried Grossmann

  • Affiliations:

  • Institute for Software Technology, Vienna University of Technology, Karlsplatz 13, A-1040 Vienna, Austria;Dept. of Statistics and Decision Support Systems, University of Vienna, Universitätsstr, 5, A-1010 Vienna, Austria

  • Venue:
  • ICSE '01 Proceedings of the 23rd International Conference on Software Engineering
  • Year:
  • 2001

Quantified Score

Hi-index 0.00

Visualization

Abstract

Defect content estimation techniques (DCETs), based on defect data from inspection, estimate the total number of defects in a document to evaluate the development process. For inspections that yield few data points DCETs reportedly underestimate the number of defects. If there is a second inspection cycle, the additional defect data is expected to increase estimation accuracy.In this paper we consider 3 scenarios to combine data sets from the inspection-reinspection process. We evaluate these approaches with data from an experiment in a university environment where 31 teams inspected and reinspected a software requirements document.Main findings of the experiment were that reinspection data improved estimation accuracy. With the best combination approach all examined estimators yielded on average estimates within 20% around the true value, all estimates stayed within 40% around the true value.