Evaluating defect estimation models with major defects

  • Authors:

  • S. Biffl

  • Affiliations:

  • Institute for software technology (E188), Vienna University of Technology, Karlsplatz 13, 1040 Vienna, Austria

  • Venue:
  • Journal of Systems and Software
  • Year:
  • 2003

Quantified Score

Hi-index 0.01

Visualization

Abstract

Based on defect data from an inspection the number of defects present in a software product can be estimated to decide on further development or quality assurance activities. Project managers are primarily interested to determine the number of major defects in the product, which may have a strong impact on product quality as well as on project schedule and cost, if they go undetected.In this work we focus on the estimation of major defects in a software requirements document. We compare the performance of objective and subjective defect content estimation techniques. For validation of the techniques we conducted a controlled experiment with 31 inspection teams which consisted of 4-6 persons.Main findings of the experiment are: (a) Objective estimation models estimated major and all defects with similar accuracy. (b) The best subjective estimations models performed for major defects comparable to the best objective estimation models. (c) The defect detection technique used influenced the set of defects found in the team, but did not make a significant difference for the accuracy of defect estimation. (d) The second inspection cycle significantly improved the estimation accuracy for all but one estimation models.