The Cost of Dichotomization

Assuming bivariate normality with correlation r, dichotomizing variable at the mean results in the reduction variance accounted for to. 647r²; and dichotomizing at the mean, to. 405r². These losses, in, result in reduction in statistical power equivalent discarding 38% and 60% of the cases under representative. As dichotomization departs from mean, the costs in variance accounted for and in are even larger. Consequences of this practice measurement applications are considered. These may not be quite so large in real data, but since are available for making use of all the original information, there is no reason to sustain.