(NCI) proposed a different approach to estimate the usual intake distribution of episodically consumed items (Tooze et al., 2006). Both methods have advantages and disadvantages.
The ISU method to estimate the distribution of episodically consumed foods assumes (sometimes unrealistically) that the probability that a person will consume the item is independent of the amount of the item consumed. This is clearly inappropriate for items such as milk and alcohol, where it is clearly seen that persons who tend to consume the items more often also tend to consume more of it when they do consume it. On the other hand, the NCI approach does not easily lend itself to implementation in nationwide food consumption surveys such as the National Health and Nutrition Examination Survey, because incorporating survey weights into the analysis requires additional programming and is quite intensive from a computational viewpoint.
Here, the committee implemented the ISU foods method and tested whether the correlation between probability of consumption and amount consumed was significantly different from zero. If not, then the committee used the estimates that are obtained from the Personal Computer-Software Intake Distribution Estimation (PC-SIDE) Version 1 when the item is declared to be a “food.” If the correlation was significantly different from zero, then the committee implemented the method suggested by Tooze et al. (2006) and then incorporated the survey weights using a parametric bootstrap approach. In most cases, the probability of consumption of the item and the amount consumed did not appear to be correlated and, therefore, the ISU foods method was deemed to be appropriate.
To estimate prevalence of inadequate intakes in a group that includes a mix of individuals of different ages who have different average requirements for the nutrient, the analysis must be performed using an approach that accounts for the differences. This situation applies to the 2–4-year and the 5–13-year age groups used for developing the recommended Meal Requirements for the Child and Adult Care Food Program (CACFP). Each of these two age groups spans two Dietary Reference Intake (DRI) age groups. The committee describes below two alternative approaches that were used to combine groups.
The first approach is appropriate when the number of persons from each DRI group in the combined group is large enough for estimation of usual intakes. The second approach, which produces prevalence estimates that closely approximate those obtained in the first approach, is more use-