All Times ET
Virtual
Beyond Unit Testing: Pseudo-Exhaustive Testing for Statistical Software (305289)
*Ryan Lekivetz, SAS / JMPJoseph Morgan, SAS / JMP
Keywords: testing, software engineering, design of experiments
For many statistical software packages, the developer and test engineer (tester) are one and the same. Usually, formal training in both software testing techniques and statistical methodology is rare. Unfortunately, the commonly used approach of unit testing, i.e. testing individual components, often misses the “corners” of the input space where bugs tend to lurk. Test engineers quickly discover that exhaustive testing is infeasible when testing statistical software, except for the simplest problems. This makes it imperative that the adopted testing approach is both efficient and effective and, at the same time, it should be based on sound principles. As it turns out, the construction of test cases can be thought of as a designed experiment (DOE). This talk provides an approach to testing based on DOE principles and shows how this approach may be applied to testing statistical software.