quantities of interest, the committee used the method of balanced repeated replication and constructed the replicated weight sets as described by Fuller (2009).The main data source for food group intake was the MyPyramid Equivalents Database, 2.0 (MPED 2.0), for U.S. Department of Agriculture (USDA) Survey Foods, 2003–2004. This database translates the amounts of foods reported in the dietary recalls from the What We Eat in America survey (USDA/ARS, 2004) into the number of equivalents for each of the 32 major food groups and subgroups in MyPyramid. The MPED 2.0 includes variables for discretionary solid fat (including some oils), added sugars, and alcoholic beverages.
Data on vitamin D were obtained from What We Eat in America, NHANES 2007–2008 (USDA/ARS, 2010a, Table 1). Limited food intake data were also available from the first Feeding Infants and Toddlers Study (FITS) (Ziegler et al., 2006). The FITS survey population included a large, representative, cross-sectional sample of parents or caregivers that provided information on the diet and eating habits of infants, toddlers, and preschoolers living in the United States. Its primary objective was to obtain information on nutrient intakes and gaps, and consumption patterns within this population. The first FITS was conducted in 2002, and the second FITS was initiated in 2008 and completed in 2009. The second survey included an expanded sample size and age range of 0 to 48 months (Nestlé Nutrition Institute, 2009), but data from the second survey were not available to the committee.
Strengths of the NHANES 2003–2004 data sets relate to the survey design, quality control, large sample size, and availability of data on nutrients of interest to the committee. Limitations relate to the lack of reliable data on supplement intake (mentioned above); the accuracy of the self-reported dietary intakes (diet recalls); and, for the purposes of this report, the relatively small number of low-income infants, children, and adults.
The under-reporting of food intakes appears to be more common than over-reporting at all age levels including children and adolescents, and especially among persons who are overweight or obese (Braam et al., 1998; Bratteby et al., 1998; Fisher et al., 2000; Little et al., 1999). There is further evidence that psychological factors contribute to under-reporting, and these may be more prevalent among obese women (Kretsch et al., 1999; Scagliusi et al., 2003). The accuracy of dietary reporting, particularly that