Abstract: An important problem in software testing is the efficient generation of test cases. Two classes of strategies, random and partition testing, have been discussed extensively in the software testing literature. In this paper, we provide a systematic statistical comparison of these two classes of strategies and demonstrate the usefulness of partition testing. We also show that some of the previous conclusions in the software testing literature about the inefficiency of partition testing are incorrect. The applicability of experimental design methods for partition testing is also discussed. A real application is used to illustrate the various concepts and to demonstrate the usefulness of experimental design methods for generating partitions.
Key words and phrases: Design of experiments, partition testing, random testing, stratified sampling, test case allocation.