Task 2: Key Concepts about Estimating Mean Nutrient Intakes from Foods and Beverages

Estimating mean intakes of selected foods and beverages is one of the most commonly conducted analyses of NHANES dietary data. 


Warning iconWARNING

In NHANES 1999-2004, the nutrient amounts in the dietary recall interview files reflect only nutrients obtained from foods and beverages, including sweetened water beverages. They DO NOT include nutrients obtained from plain drinking water. Beginning in 2005, nutrients from plain drinking water will be included in the data release.

To obtain complete and accurate results from your analysis, consider the following issues before you begin.

Addressing Random Error and Bias

Nutrients can be obtained from diet (i.e., from foods and beverages) or from dietary supplements.  Sometimes analysts are interested in examining nutrient intake from diet, sometimes from supplements, and sometimes from both.  This task relates to estimating nutrient intake from diet only.  That is, these estimates do not represent total nutrient intake from all sources.

Nutrient intakes from foods and beverages are estimated using dietary recall data.  Although dietary recall data are known to contain random errors, especially large day-to-day variability, as noted in the previous module, we typically assume these errors cancel out when estimating means.  Therefore, no specific statistical adjustment is necessary, and the mean of 1-day intakes can be used as an estimate of the mean of the usual intake distribution in the population.

As noted in the last task, dietary recall data are also known to contain bias, at least insofar as a tendency toward underreporting of energy.  Little is known regarding the extent to which energy underreporting extends to underreporting of other nutrients.  For that reason, and for practical purposes, the current statistical convention is to assume that the recalls are not biased (i.e., that no underreporting occurs).  However, this assumption is more troubling than the one regarding random error and should be noted as a caveat in an analysis of this type.


Interpreting Measures of Central Tendency

If the data are highly skewed, as dietary data often are, means may not provide a very good representation of central tendency.  You may want to consider using the median instead of, or in addition to, the mean in such an instance.  However, you should know that the simple median of reported intakes from recalls is not clearly interpretable with regard to usual intake (as it really represents the median on a given day).  Information on how to obtain the distribution of usual intake and its associated median can be found in the Advanced Dietary Analyses course.


Using Appropriate Statistical Procedures

Also, as mentioned in the last task, the standard errors of estimated means should be reported, to provide an indication of the variation about the mean.  Special statistical procedures are required to get appropriate standard errors when using data from a complex sample such as the NHANES.  In addition, the appropriate sample weights should be applied because the inference should be to the population rather than the sample.  See “Module 13: Estimate Variance, Analyze Subgroups, and Calculate Degrees of Freedom” for more information.



When estimating the mean of the population distribution of usual dietary intakes from 24-hour recalls, single day data are sufficient and no specific statistical adjustment is necessary, but an assumption regarding lack of bias is required and should be acknowledged. The second day of dietary recall is generally not used to estimate means but is used for more advanced analyses.


close window icon Close Window to return to module page.