Estimation of population distributions of habitual nutrient intake based on a short-run weighed food diary.

There have been many attempts to characterize day-to-day variation in nutrient intake. This variation has a fixed component, associated with particular days of the week, and a random component. Both components were studied for a range of nutrients, using 4 d weighed diary data from a large, nationally representative survey of people aged 65 years or over. Since day-to-day variation may distort the characterization of the population distribution of habitual nutrient intakes, especially when diets are studied over only a small number of days, a statistical method was developed to correct for this distortion. Results suggested that population distributions of habitual nutrient intake could be accurately constructed from 4 d weighed diary data and that the method might be successfully applied to studies based on as little as 2 d of observation. The method is particularly valuable for correcting estimates of extreme nutrient intakes for biases induced by uneven representation of days of the week and by within-person variation.