A revisit of the proportional sampling strategy

When test cases are selected with replacement, we have P/sub pss//spl ges/P/sub r/, where P/sub pss/ and P/sub r/ are the probabilities of detecting at least one failure by the proportional sampling strategy and random testing respectively. However, recent results on the proportional sampling strategy have highlighted that P/sub pss//spl ges/P/sub r/ may be very sensitive to whether test cases are selected with or without replacement. The paper provides more insights into this important issue, as the proportional sampling strategy is the most practically applicable partition testing strategy and it always performs better than random testing.