[Skip to Navigation]
December 1997

Another View on the 'Right' Statistical Measure of Twin Concordance

Arch Gen Psychiatry. 1997;54(12):1126-1128. doi:10.1001/archpsyc.1997.01830240086012

THIS ISSUE of the ARCHIVES includes a paper by Chmura Kraemer1 advocating the use of the K statistic and questioning the use of the tetrachoric correlation coeffiecient (TCC) in twin research in general and in our recent twin study2 in particular. Falconer3 suggested that many traits of biological interest are inherited multifactorially, but distributed discontinuously; the underlying continuum is related to the visible scale by the threshold on the continuous scale that corresponds to the discontinuity on the visible scale. If this underlying variation could be measured directly, it is assumed it would be normally distributed. The TCC is an estimate of the Pearson correlation coefficient between the latent continua. Large discrete genetic or environmental effects will falsify the assumption of bivariate normality as will certain kinds of genotype-environment interactions. The assumption of bivariate normality, however, cannot be tested directly with dichotomous items, which means that we

Add or change institution