Jaymie Strecker, Affinities Between Test Suites and Faults
When designing a test suite, testers must compromise between the reliability of the tested software and the cost of testing, yet the testing literature offers little guidance. So far, empirical studies have compared the numbers of faults that different testing techniques detect in "realistic" programs that contain "realistic" faults. Given a test suite for a different program, however, these studies can predict neither how many nor what kinds of faults the test suite will find. In this study, we explore the relationship between test suites and the faults they detect. From an analysis of test-suite and fault characteristics in several moderate-sized, GUI-based programs, we derive some statistical models of how these characteristics affect fault detection.