An official website of the United States government
How to Choose an Analysis Method for Monitoring Dietary Intakes
When choosing an analysis method for surveillance, consider the purpose, available data, and developed methods. For example, there may be interest in estimating the mean and the distribution (e.g., percentiles) of intake for a population or subpopulation. Also, several methods have been developed for use with 24-hour recall (24HR) data for this purpose (FFQs are not recommended for surveillance research). The table below provides guidance, considerations, and caveats.
Statistical Analysis Methods: Describing Dietary Intake for a Population/Group1
| Purpose | Minimum Number of 24 HRs | Method | Recommendations | Considerations & Caveats |
|---|---|---|---|---|
| To calculate mean but not distributions, of both total and component HEI scores | 1 | Simple HEI Scoring Algorithm | Possible | Does not estimate usual intake |
| Mean Ratio Method | Possible | |||
| Population Ratio Method | Recommended | Means calculated with the population ratio method are slightly more biased than the bivariate approach, usually, but not always, in the direction of overestimation | ||
| To calculate mean and distribution of individual component scores; mean, but not distribution, of total HEI score (by summing components) | 2 (on at least a subset) |
Bivariate Approach | Recommended Markov Chain Monte Carlo (MCMC) also recommended |
Doesn’t account for relationship between HEI components (Does NOT provide distribution of the total HEI score) Gives estimates of means consistent with the MCMC method More user-friendly than MCMC method |
| To calculate mean and distribution of both total and component HEI scores | 2 (on at least a subset) |
Multivariate Approach (MCMC) | Recommended | Provides best estimates of mean and distribution; least biased method because it accounts for the relationship between HEI components Code is more computationally intensive than any other method and has a steep learning curve. |
1 Freedman et al. (2008) examined the simple HEI scoring method, the mean ratio method, and the population ratio method and found that, in general, scores calculated using the population ratio method were less biased than those obtained using the other two methods. Ideally, the HEI would be calculated on the basis of usual intake, which requires 2 or more 24HRs, followed by use of a statistical method to estimate usual intake.