Adjustment cycles

Previous  Top  Next

The FACTORFINDER-Procedure adjusts the counter matrix in revolving cycles. An adaption cycle consists of

1. generating a tolerance test vector and

2. evaluating it based on comparisons of candidate example and reference examples and then

3. changing the counter values depending on the result of the evaluation.

The counter matrix receives its Gestalt by varying combinations of factors in which candidate example and reference examples differ. The tolerance derived from one comparison is considered valid for all the other comparisons. This allows to generalize and to fuse all tests results. If the attribute "tolerance" were a property of tolerance-confirming example-pairs, then tolerance-expectations could not be transferred from pair to pair and the information of the counter matrix would be worthless.
How much a candidate deviates from ideal can not be inferred from a single comparison. Solely after many successful reference examples deviate subcritically from the same ideal conclusions become feasible. As the ideal values never are known with certainty, the successful reference examples serve as a surrogate. Altogether the analyst gains an overview on what factors are important. From the relative position of each alternative inferences on the relevance are drawn. They describe tendencies that contribute to the understanding of the object analysed.
A candidate example that is close to failure and just acceptable makes all factors look more critical than an example with only ideal values. Consequently, there are as many different perspectives on relevance as there are successful examples. The evaluation shows how tolerable ideal values are from the perspective of the current candidate example.