9i
a) In detective test, the method of error generation is a powerful
device for the judgment. Known errors are introduced or gener
ated artificially as stimuli for distortions. From the recorded dis
tortions, the errors are derived in the data processing stage, and
compared with the errors originally introduced.
The obtained mean discrepancy and the spread of individual
values around it represent the estimates for the bias and sharpness
of the testing method. This kind of estimation applies to the
adequacy of the test. Its main limitation lies in the possibility for
the generating of artificial errors.
b) In detective and comparative tests, the method of repeated test
runs is frequently applied. This method is based on the assump
tion that the test runs, repeated under similar conditions,
should provide essentially the same results (estimates). Thus,
this method applies to the repeatability.
The method is efficient only if significant errors (disturbances)
are present. The consistency and repeatability cannot be esti
mated if the test runs are applied to nearly perfect systems. In
systems with significant disturbances, the repeatability of a
testing method can be estimated from the distribution of the
results of repeated test runs.
A combined analyticalexperimental approach appears to
be most convenient for the judgment of the power of detective
tests.
c) In comparative tests, intended for the judgment of the overall
performance, the system operates in representativeincluding
extremeconditions. The criteria for the estimation of the po
wer are therefore different from the case of detective tests.
The power of comparative tests lies in their broadness and
consistency
4.7. Optimisation
The optimisation of a testing process is based on the evaluation
of the power of the test and the subsequent analysis and synthesis.
The evaluation is in fact a comparison of the achieved and desired
power of a testing process. If the achieved power is not satisfactory,
the testing process (or the desired powermust be revised.
The analysis applies to the detection of the weak links in a testing
process. For this purpose, additional experiments are sometimes
necessary. After detecting weak links, a refined synthesis can be
set up.
In detective tests the refinements apply to the response of the
system to relevant errors. They concern the standard input (struc
ture, sample size), auxiliary devices, structure of the testing pro
cess and the corresponding mathematical model, measuring tech
nique and data processing.