Document Type



The correspondence of marker gene and anthropometric data has been tested for seven villages in four tribes of South American Indians. General­ized distances calculated from each kind of multivariate data have been used to construct dendrograms which represent the relationships or cluster-structure implicit in the two kinds of data. For each set of data, the “best” 50 dendrograms are identified by the criterion of minimum string or total path length. Thirty-one dendrograms are found in common among the best 50, indicating statistically highly significant correspondence by comparison with pairs of dendrograms randomly generated by a computer simulation. The effectiveness of the angular transformation, used in the calculation of genetic distances to stabilize the sampling variance of allele frequencies, has been tested here by computer simulation. For samples of size 60 (30 individuals) or larger, the binomial variance is stabilized adequately over the allele frequency range 0.005 to 0.995. It appears that standard practice restricting the use of this transformation to the range 0.05 to 0.95 is unnecessarily conservative.