Cover Image


View/Hide Left Panel

Box D-1 Major findings—Intakes and requirements are correlated

  • When the SDr is small relative to the SDi, no serious biases on the estimate of prevalence are evident even at correlation values as high as 0.5 or 0.6 (Figure D-1 and Figure D-4).

  • When the SDr increases relative to the SDi, increasing the correlation between intakes and requirement can result in noticeable biases in the prevalence of inadequacy even when the correlation is no larger than about 0.4 (Figure D-2 and Figure D-5).

  • When the SDr is as large as the SDi, the bias in the estimate of prevalence can be significant even if the correlation between intakes and requirements is 0. This indicates that the EAR cut-point method is less robust to departures from the last assumption (variance of requirements must be smaller than variance of usual intake) (Figure D-3 and Figure D-6).

  • When mean intake is equal to the EAR (prevalence is exactly equal to 50 percent), neither increasing the correlation coefficient to 1 nor equating the variances of requirements and intakes introduces a bias in the estimated prevalence (Figure D-7, Figure D-8, and Figure D-9).

FIGURE D-1 The effect of correlation between usual intake and requirement on the prevalence of inadequate intakes estimated using the Estimated Average Requirement (EAR) cut-point method for 10 values of the correlation. For all correlations, mean intake = 90, standard deviation (SD) of intake = 30, EAR = 55, and SD of requirement = 7.5 units.

NOTE: When the SD of requirement is small relative to the SD of intake, there is no serious bias of the EAR cut-point method until correlation reaches 0.5 to 0.6.

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement