A good
calibration of a relationship results in a random sequence of the positive and
negative residuals of prediction. Calibrations with systematic errors show longer
sequences of positive and negative residuals. For example, in figure
29 the mean residuals of R22 show only 3 sequences, first with negative
signs, then with positive signs and finally with negative signs again. These
sequences are also known as runs. The Wald-Wolfowitz method tests whether the
number of runs is small enough or big enough for the null hypothesis of a random
distribution of the signs to be rejected [232].
Thereby the number of positive and negative runs is compared with the tabulated
value for the number of observations and for a given error level [233].