Fault Detection Effectiveness of Spathic Test Data

• Authors: Jane Huffman Hayes;Pifu Zhang
• Venue: ICECCS '02 Proceedings of the Eighth International Conference on Engineering of Complex Computer Systems
• Year: 2002

This paper presents an approach for generating test datafor unit-level, and possibly integration-level, testing basedon sampling over intervals of the input probabilitydistribution, i.e., one that has been divided or layeredaccording to criteria. Our approach is termed "spathic"as it selects random values felt to be most likely or leastlikely to occur from a segmented input probabilitydistribution. Also, it allows the layers to be furthersegmented if additional test data is required later in the testcycle.The spathic approach finds a middle ground between themore difficult to achieve adequacy criteria and random testdata generation, and requires less effort on the part of thetester. It can be viewed as guided random testing, with thetester specifying some information about expected input.The spathic test data generation approach can be used toaugment "intelligent" manual unit-level testing. An initialcase study suggests that spathic test sets detect more faultsthan random test data sets, and achieve higher levels ofstatement and branch coverage.