CASE FOR NONPARAMETRIC STATISTICS IN ANALYSIS OF TWIN DATA

The distribution of within pair differences in measurements of twins used in studies of heritability often show considerable skewness. If the skewness is severe, the commonly used statistics for evaluation of heritability, the F test or HOLZINGER’s h2, may be inappropriate. The shape of distributions of differences in monozygous twin measurements depends on environmental factors as well as errors in measurement. Differences in measurements between dizygous twins depend on the above plus hereditary factors such as phenotypic frequency, chromosomes in common, assortative mating, and differential interaction between heredity and environment. Evidence in the form of the probabilities from Chi Square tests, F ratios, and Mann-Whitncy I tests made on the same data, demonstrates that a nonparamctric statistic such as the Mann-Whitney U test may be more appropriate for the analysis of twin differences. © 1969 S. Karger AG, Basel.