Substituting Voas´s testability measure for Musa´s fault exposure ratio

This paper analyzes the relationship between Voas´s (1991) software testability measure and Musa´s (1987) fault exposure ratio, K. It has come to our attention that industrial users of Musa´s model employ his published, experimental value for K, 4.2×10^-7, for their projects, instead of creating their own K estimate for the system whose reliability is being assessed. We provide a theoretical argument for how a slight modification to Voas´s formulae for determining a predicted minimal fault size can provide a predicted average fault size. The predicted average fault size can be used as a substitute for 4.2×10^-7, and in our opinion is a more plausible choice for K