FIGURE D-19 The effect of the skewness of the requirement distribution on bias of the estimated prevalence of inadequate intakes using the Estimated Average Requirement (EAR) cut-point method for five values of skewness. For all levels of skewness, mean intake = 12 mg, standard deviation (SD) of intake = 3 mg, and correlation between intake and requirement = 0. The SD of requirement varied with the skewness of the requirement distribution.

no information was available to indicate nonsymmetrical distributions of requirements, so symmetry was assumed for the nutrients studied (IOM, 1997, 1998b, 2000).

When requirements are not symmetrically distributed around the EAR, the probability approach should be used to assess prevalence of inadequacy. To implement the probability approach it is necessary to specify a probability model for the requirement distribution. The probability approach should result in essentially unbiased estimates of prevalence if a skewed requirement distribution is accurately specified. If the requirement distribution is incorrectly specified (for example, a log-normal model is chosen for estimation, but gamma or Weibull would be more correct), then the prevalence estimates obtained via the probability approach will also be biased. The effect of incorrect model specification on the bias of the probability approach has not been studied, but the bias resulting in this case would likely still be smaller than that resulting from the application of the EAR cut-point method to estimate prevalence.