. "Appendix D: Assessing the Performance of the EAR Cut-Point Method for Estimating Prevalence." Dietary Reference Intakes: Applications in Dietary Assessment. Washington, DC: The National Academies Press, 2000.
The following HTML text is provided to enhance online
readability. Many aspects of typography translate only awkwardly to HTML.
Please use the page image
as the authoritative form to ensure accuracy.
DRI DIETARY REFERENCE INTAKES: Applications in Dietary Assessment
In summary, violating the independence assumption (i.e., a non-zero correlation) is likely to produce relatively minor biases on the estimates of prevalence obtained from applying the EAR cut-point method as long as the correlation between intakes and requirements does not exceed 0.5 or 0.6; the SDr is substantially smaller than the SDi; and the true prevalence is neither very small nor very large. The use of the EAR cut-point method (or the probability approach) is not recommended for investigating the adequacy of energy intakes in any group because for food energy the correlation between intakes and requirements is known to be very high.
VARIANCE OF REQUIREMENTS IS LARGE RELATIVE TO VARIANCE OF INTAKES
To test the effect of violating the assumption that variance of requirements must be substantially smaller than variance of intakes for good performance of the Estimated Average Requirement (EAR) cut-point method, various scenarios were considered. Mean intake was fixed at 90 units and SDi at 30 units, as before, and 0.01 and 0.7 were chosen for the correlation between intakes and requirements. The EAR was fixed at three different values: 55, 70, and 90 units. For each of the six different scenarios, the SDr varied from a low value of 0 to a high value of 40 units, in 5 unit increments.
Again, for each case, a large population was generated, and groups of 2,000 individuals were sampled 200 times. The prevalence estimates shown in each case are obtained as the average over the 200 replicates.
Box D-2Major findings—Variance of requirement relative to variance of intake
The impact of increasing the SDr relative to the SDi on the bias of the prevalence estimates can be large, especially when true prevalence is not close to 50 percent (Figure D-13 and Figure D-15).
When the correlation between intake and requirement is high (0.7), the bias in the estimated prevalence can be high, but it does not increase monotonically as SDrincreases (Figure D-14 and Figure D-16).
When true prevalence is 50 percent, increasing the SDr even to values above the SDi has no impact on the estimates of prevalence.