查看文章

A little further analysis of the data em-ployed by both Vining and me reveals, for the aggregate, that the variance is. 62 for logA and logD and. 41 for logP, with a covariance of. 21 between logA and logP. Consider Vining's interpretations above in the light of these results. As Vining suggests, the observed clustering of slopes around-2/3 could simply mean that the variance of logA is gen-erally twice that of logP and nothing more; empirically, however, the ratio is 3 to 2, not 2 to 1. The expression for the expected slope between logA and logD would be that given by Vining if the covariance between logA and logP were zero; empirically, however, it is not zero. The slope relating logA and logP would be-2 if the-2/3 slope between logA and logD represented a case of perfect correlation; empirically, however, the correlation between logA and logD is far from perfect (r2=. 42), and it is thus not surprising that the slope …