2022 年 130 巻 2 号 p. 75-91
Although studies of population affinity based on nonmetric traits have achieved remarkable successes, most of these studies seem to present a methodological problem. Since evidence indicates a threshold model for most nonmetric traits, the use of individual counts in studies would seem to have wasted a considerable amount of useful information to produce less reliable results. A review of relevant articles suggests that the use of this baseless methodology has persisted by neglecting an inconvenient truth. To improve the situation, the author proposes a generalized theory based on the assumption of constant within-individual instabilities, which covers both the standard threshold model and the single-genotype model. The proposed theory proves the general validity of Ossenberg’s proposals, i.e. the use of side counts for threshold traits and an examination of etiology by correlating the proportion of asymmetry and the trait frequency. The data of 28 nonmetric traits collected by Ossenberg were examined using the theory. The proportion of asymmetry was negatively correlated with side counts in all the traits with statistical significance. The threshold model exhibited higher goodness of fit than the single-genotype model for 25 traits. The loss of information caused by using individual counts for threshold traits instead of side counts is estimated to be equivalent to a considerable decrease (16–40%) in sample size. The use of both sides improves the reliability of the tetrachoric estimation of inter-trait correlation comparable to a 1.6- to 2.6-fold increase in the sample size by enabling the use of their four combinations. It was also shown that the theory makes it possible to estimate the penetrance rate of congenital anomalies and tumors from the proportion of asymmetry.
