The estimation of reliability of modules using code coverage is motivated by a two-phase method for estimating the quality of software using static structure and predicted quality estimates of component modules. A pilot experimentwasconducted to investigatethe accuracy of reliability estimates obtained from code coverage. The code coverage was measured using random testing. Three parameters that were varied are the size of programs, variations in operational prole, and the fault density. Results fromthis experiment were analyzed to understand how the accuracy of reliability estimates is aected by: (1) variation in fault density and (2) the size of programs. The results indicate that the degree of correlation between coverage and reliability has an inverse relationship with the fault density, and a direct relationship with program size.
Saileshwar Krishnamurthy, Aditya P. Mathur