Today we continue to discuss software testing effectiveness measures - these are  P-Measure, E-Measure and F-measure. We were discussing random testing. The P-measure in this case is equal to 1 minus the success rate raised to the power of the inverse of failure rate p. The E-measure is the combination of the number of test cases and the failure rate. The F-measure is the ratio of the size of the input domain and the size of the failure region. Random black box sampling of the input domain gives the worst case. That said the randomness may still be required because the failure patterns are not always known. This leads to a technique called adaptive random testing. This is  a technique to distribute test cases evenly within the input space without losing all randomness. The fixed size candidate set case of this technique involves the following steps; First, randomly generate a set of test cases from each iteration of the test case generation. For each such set, the distance of the candidate from the set of previously executed tests is calculated. Lastly, the candidate with the greatest distance is executed and added to the set of executed tests. This technique has achieved result near the upper bound for effectiveness.
In subdomain testing, the effectiveness of testing is measured in terms of the faults detected and the maximum value of E-measure gives such a case. In contrast, to maximize P-measure, it is only necessary to have one subdomain full of failure causing inputs, even though all other subdomains may have low failure rates. Therefore the E-measure has more merit for this purpose. Besides, it can distinguish the capability of detecting more than one failure while the P-measure treats them the same if they result in at least one failure.  Furthermore, it has been difficult to extend the formal analysis to more general and more common case of overlapping subdomains. By using E-measure, T.Y.Chen and Y.T.Yu say that new properties of P-measure can be found rather than with P-measure alone. For example when comparing subdomain testing with random testing, a greater value or equal value of  E-measure assures a greater value of P-measure. Thus the relative value of E-measure also gives a good indication of P-measure.
There is no single measure that serves as the best measure.