is also needed in order to estimate the prevalence of inadequacy.

In addition, it was stated that the estimates of inadequacy would be essentially unbiased when the actual prevalence of inadequacy in the group is close to 50 percent. As the true prevalence approaches 0 or 100 percent, the performance of the EAR cut-point method declines, even if the conditions listed above are met.

To test the EAR cut-point method, some preliminary simulation studies were performed. The reliability of this method of estimating the prevalence of inadequacy was evaluated in cases where the assumptions above were met, and also in cases in which one or more of the assumptions were violated. For example, the EAR cut-point method was used to evaluate groups in which (1) intakes and requirements were correlated (for example, food energy), (2) the standard deviation of requirements (SD_r) was larger than the standard deviation of usual intakes (SD_i), and (3) the distribution of requirements was skewed (as is the case of iron in menstruating women).

This appendix does not test the performance of the probability approach. The probability approach, by construction, will perform well whenever intakes and requirements are independent, and whenever the form of the distribution of requirements is known. As in the EAR cut-point method, a reliable estimate of the distribution of usual intakes in the group must be available to ensure an unbiased estimate of the prevalence of inadequacy in the group.

Results of the simulation studies are reported in three sections. The first section examines the impact of violating the independence assumption on the estimates of prevalence. In the second section, the robustness of the EAR cut-point method to departures from the assumption of small SD_r relative to SD_i is tested. Finally, in the third section, the effects of departures from the assumption of a symmetrical requirement distribution are considered. In each section, a description of how the simulations were run is followed by a summary of the major findings. The simulation studies presented are preliminary and by no means definitive. They are intended to provide initial insight into the performance of this short-cut of the probability approach for estimating inadequacy. It is hoped that this report will encourage other researchers to proceed from the information presented here and conduct further research on this important topic.

The impact of violating the assumption of independence between intakes and requirements was evaluated by estimating prevalence of