High counter values represent not only a specific evaluation of the tolerance. Additionally, they deliver an estimate on the validity of that evaluation. To put is shortly: When evaluating the counter matrix two threshold values come into play. The first threshold value is set initially by the analyst. It is the minimum counter value required to have a well-founded tolerance hypothesis. The second threshold value is found during analysis. This value marks the border between the well-founded tolerance hypothesis and those values that are represented only by little counter values.
As soon as the preset mimimum number of adaption cycles is reached, the procedure verifies whether the counter matrix shows a mature distribution of counter values. In this stadium, the tolerance options beginning with 0.0 up to a certain value are well-founded and therefore represented by a counter value above the threshold value. All other tolerance options larger than this value are represented by small counter values. As long as there is no obvious border between those two intervals, repeatedly new adaption cycles are initiated. Stable distributions develop in each dimension only if the other dimensions develop equally stable distributions. By that, it is possible to judge the reliability of the analysis.
The tolerance options represent the tolerability of non-ideal values. They depend on how many reference examples differ from the candidate example in the respective dimension. At the same time, the adequacy of the tolerance hypothesis depends on the tolerance hypothesis assigned to all the other factors. The higher tolerance hypothesis in other factors are, the less high tolerance hypothesis in the current dimension will find confirmation. This is crucial to represent the interdependence of all dimensions. Those interdependencies decide what tolerance options get selected most likely into the tolerance test vector.
The counter matrix above shows in dimension 1 well-founded tolerance options up to the value of 0.2. In dimension 2, the upper limit of well-founded tolerance options is 0.6, in dimension 3 it is 0.9. Dimension 4 has the same upper limit of 0.9. In dimension 5, all tolerance options up to the value 0.7 are represented by high counter values.