A Statistical Approach for Improving the Performance of a Testing Methodology for Measurement Software.

This paper describes the significant enhancements brought to an original methodology designed for testing measurement software. In a previous paper, the authors proposed a black- box seven-step procedure that allows the functional verification of complex instrument software to be performed. The main features of the procedure are concerned with the following: 1) the ability of reproducing actual correlations among the software inputs and 2) the need for a limited number of test cases. Making use of innovative statistical techniques, the methodology performance and reliability have been enhanced. Two further steps have been added with the aim of improving the correlation coefficient assessments and providing the estimations with a confidence level. Finally, a new strategy has been studied to optimize the number of test cases. The effects of the new solutions on the performance of the methodology are evaluated by applying the procedure to a complex software module employed in an automotive system. A comparison with the previous methodology version is also reported. [ABSTRACT FROM AUTHOR]