The values of P(n, k) and of the cumulative probabilities
543
i=O are displayed for selected values of n and k in Table 1.11 The concrete process for applying this test is then as follows: (a) we fit least squares regression lines to the data according to both the linear and ratio models; (b) we order the absolute values of the residuals from each regression such that the number of peaks will tend to be large if the hypothesis employed in fitting the particular relation (linear or ratio) is false;12 (c) we count the number of peaks in both series of residuals and compare them with Table 1; de- pending on the outcome of the comparisons we accept one or the other or neither hypothesis analogously with the procedure discussed under the F-test.