On random and partition testing

There have been many comparisons of random and partition testing. Proportional partition testing has been suggested as the optimum way to perform partition testing. In this paper we show that this might not be so and discuss some of the problems with previous studies. We look at the expected cost of failures as a way to evaluate the effectiveness of testing strategies and use it to compare random testing, uniform partition testing and proportional partition testing. Also, we introduce partition testing strategies that try to take the cost of failures into account and present some results on their effectiveness.