Capture-recapture models help software managers by providing post-inspection defect estimate remaining in a software artifact to determine if a re-inspection in necessary. These estimates are calculated using the number of unique faults per inspector and the overlap of faults found by inspectors during an inspection cycle. A common belief is that the accuracy of the capture-recapture estimates improves with the inspection team size. This however, has not been empirically studied. This paper empirically investigates the effect of the number of inspectors on the estimates produced by capture-recapture models, by using inspection data with varying number and types of inspectors. The results show that the SC (Sample Coverage) estimators are best suited to software inspections and need least number of inspectors to achieve accurate and precise estimates. Our results also provide a detailed analysis of the number of inspectors necessary to obtain estimates within 5--20% of the actual defect count.
展开▼
