Estimating the Required Code Inspection Team Size

Abstract

Code inspection is considered an efficient method for detecting faults in software code documents. The number of faults not detected by inspection should be small. Several methods have been suggested for estimating the number of undetected faults. These methods include the fault injection method that is considered to be quite laborious, capture recapture methods that avoid the problems of code injection and the Detection Profile Method for cases where capture recapture methods do not provide sufficient accuracy. The Kantorowitz estimator is based on a probabilistic model of the inspection process and enables the estimating the number of inspectors required to detect a specified fraction of all the faults of a document as well as the number of undetected faults. This estimator has proven to be satisfactory in inspection of user requirements documents. The experiments reported in this study suggest that it is also useful for code inspection.