Because it is impractical to provide DVs on the nutrition label for each subpopulation, it is necessary to compute a single number that will summarize the distribution of requirements in the total population. The logic described above argues in favor of choosing a central value of the distribution as the DV. For symmetrical distributions, such as the normal distribution, the mean and the median are identical. However the distribution of requirements for the population, derived from the distributions for the subpopulations, in general will not be symmetrical (see Chapter 4). Therefore the median, with 50 percent of the requirements above and 50 percent below, is preferred to the mean, which is sensitive to extreme values of requirements. In summary, the DV should be defined as the median of the population distribution of requirements. This is represented by the population-weighted EAR for nutrients where the distribution of requirements is known. Derivation of this value takes into account the relative proportions of the population in each of the 13 life stage and gender groups that comprise the target population for the Nutrition Facts box and the EAR and the CV of the requirement distributions for each group.

To compute the population distribution of requirements for the DV, the subpopulation distributions are combined using weights obtained from census data. The DV is the median of this resulting distribution. This procedure is easily adapted for different demographic profiles, such as for the Canadian population or for different projected future populations (see Appendix B).

Specifically, to calculate the population-weighted EAR for each subpopulation defined by life stage and gender, the requirement for each nutrient is assumed to have a distribution. For nutrients having an EAR, this distribution is assumed to be normal with the median equal to the EAR and a CV of 10 percent. Two exceptions are vitamin A and niacin, which have assumed CVs of 20 percent and 15 percent, respectively. The following text illustrates how the weighting could be approached for nutrients with CVs equal to 10 percent. Slight modifications are required for the two exceptions.

As an example, let the population of interest be females and males ages 4 years and older (excluding pregnant and lactating females) in the United States. As stated earlier there are 13 subpopulations with EARs in this population: all children ages 4 to 8 years, and for males and females, separate groups based on the following age breaks: 9 to 13 years, 14 to 18 years, 19 to 30 years, 31 to 50 years, 51