Measurement of Relative Variation: Sociological Examples

There exists a variety of situations in which the dispersion of data, rather than their mean or other central value, is of interest. The coefficient of variation, which is a measure of dispersion divided by the appropriate measure of central tendency, is preferable to a raw measure of dispersion for this purpose. However, these measures do not have a constant range, and the common practice of dividing the variation value obtained by 100 and expressing the result as a per cent is particularly inappropriate, as all have maxima larger than unity for n>2. The present paper provides standardization procedures for coefficients of variation. The resulting standardized coefficients (called S-measures) have a zero-to-one range, and the common practice of dividing the variation value obtained by 100 and deviation) for descriptive purposes is demonstrated. It is suggested that S-measures replace, or at least supplement, coefficients of variation and raw dispersion measures when the problem of interest is relative dispersion between groups.