Current concerns over links between diet and health, the safety of food, and developing effective nutrition education and food assistance programs have placed increasing demands on federal data collection and methods of monitoring the food supply.

Design of effective food and nutrition policies, efficient allocation of resources, and more precise targeting of programs require good estimates of the percentage of the population with deficient, or excess, nutrient or other food component intake. An individual's mean daily intake of the dietary component is a good estimate of the individual's dietary status. However, to evaluate dietary adequacy of a population it is necessary to obtain an estimate of the distribution of usual intakes. Often, the distribution of usual intakes is estimated from the distribution of mean daily intakes. Further, it is usually assumed that usual intakes of nutrients are normally distributed.

Two problems arise. First, distributions of usual intakes for most nutrients are skewed. Second, the variance of the distribution of mean daily intakes is larger than the variance of the true usual intake distribution, due to within-individual variability of daily intakes. Some proposed adjustments produce the correct mean and variance in the estimated distribution, but fail to correct the skewness, unless the true distribution of usual intakes is normal.

We describe a method for estimating usual intake distributions which does not assume normality, and which takes into account the within-individual variation in daily intake. The method consists in transforming the dietary data from the original space into normal space, and predicting the usual daily intakes in normal space. Inferences about the dietary status of the population can then be made in normal space. Alternatively, predicted normal usual intakes can be transformed back to obtain a set of pseudo usual intakes in the original scale.