Cover Image


View/Hide Left Panel

intake distributions, statistical procedures that assume that nutrient intake data are normally distributed cannot be applied to these data.

Day-to-Day Correlation in Intake Data Collected over Consecutive Days

When intake data are collected over consecutive days, observations for an individual cannot be assumed to be independent because what is consumed on one day often affects what is consumed on the next. This effect can work several ways—the same meal may be repeated the next day (as with leftovers) or the same food may be avoided two days in a row (as with liver). In either case, the assumption of independence for within-person observations does not hold unless dietary intake data are collected several days apart. The length of time needed between observations so that independence can be assumed depends on the dietary component. For energy, for example, it suffices to space daily observations one or two days apart, but for vitamin A, which is not present in all foods, a three- to four-day gap between 24-hour recalls for the same individual might be necessary to guarantee independence among observations.

Other Survey-Related or Nuisance Effects

Dietary intake data are often collected in nationwide food consumption surveys that have a complex design and response rates under 100 percent. In these cases, each respondent carries a sampling weight that corrects that individual's importance in the sample. These weights must be carried throughout the procedure for estimating usual intake distributions if this estimated distribution is to be used to make inferences about the wider population from which the group was drawn.

Overview of Methods to Adjust Mean Intake Distributions

Because of the above attributes of dietary intake data, obtaining reliable estimates of usual intake distributions is not straightforward. The NRC, in its 1986 report, set forth the concept of a usual intake distribution, and proposed a statistical approach to adjust observed mean intake distributions to partially remove the day-to-day variability in intakes. The resulting estimated usual intake distribution has a spread that approximately reflects the between-individual variability in intakes (NRC, 1986). Aickin and Ritenbaugh (1991) pro-

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