We propose tuple density to be a new metric for combinatorial test suites. It can be used to distinguish one test suite from another even if they have the same size and strength. Moreover, it is also illustrated how a given test suite can be optimized based on this metric. The initial experimental results are encouraging.

Combinatorial testing is an important approach to detecting interaction errors for a system with several parameters. Existing research in this area assumes that all parameters of the system under test are always effective. However, in many realistic applications, there may exist some parameters that can disable other parameters in certain conditions. These parameters are called shielding parameters. Shielding parameters make test cases generated by the existing test model, which uses the Mixed Covering Array (MCA), fail in exposing some potential errors that should be detected. In this paper, the Mixed Covering Array with Shielding parameters (MCAS) is proposed to describe such problems. Then test cases can be generated by constructing MCAS's in three different approaches. According to the experimental results, our test model can generate satisfactory test cases for combinatorial testing with shielding parameters.