In their interesting paper Cohen et al. [1972] analyze results of a large quantity of painstaking observations and make a determined effort to explain their observations physically in terms of vertical mixing in the atmospheric boundary layer. Much of their discussion is based on conclusions drawn from visual inspection of scatter diagrams, their Figures 6, 7, and 8. In particular, they say. with regard to Figure 6, 'groups i and 2 . . . are not clearly distinguishable by [AT],' and that 'groups 1 and 2 are not distinguishable [in Figure 7] on the basis of SR6. On the other hand, SR3, shown on Figure 8, does separate group 1 from group 2.' They interpret this observation in terms of the vertical scale of mixing. The differences between the figures do not seem to be a sufficiently sound basis for these statements; therefore I have attempted to test whether these representations of the data are significantly different.In testing the hypothesis that two sets of numbers are samples from the same population the Wilcoxon (Mann and Whitney) statistic is useful, because it is easy to compute and its • Contribution 130, use is rather free of implied assumptions. As a measure of how well Figures 6, 7, and 8 Figure 7 and four from Figure 8. The sums of ranks for group 2 in Figures 6, 7, and 8 are 448, 471, and 473 (___2, approximately), respectively. The improvement due to using wind speed in Figures 7 and 8 is roughly equivalent to moving two points from one end of the 'cloud of points' to the other, or, alternatively, changing all the ranks of one of the sets by 1. Including the wind speed is physically appealing but does not seem to be adequately supported by the observations. The data surely do not allow discrimination between stability ratios for deeper and shallower layers as predictors for the vertical gradient of radon concentration.
distinguish between groups i and 2 we can use the value of that statistic, or, equivalently, the T (summed rank) statistic. For this purpose (and for any approximately equivalent subjective analysis) it is necessary to have the same number of observations for the different figures; this was achieved by omitting central points, two from
t•EFERENCECohen, L. D., S. Barr, R. Krablin, and H. Newstein, Steady-state vertical turbulent diffusion of radon, J. Geophys. Res.,
