Variability of food intakes. An analysis of a 12-day data series using persistence measures

Data from the US Department of Agriculture's Exploratory Study of Longitudinal Measures of Individual Food Intake, conducted in 1982, were used to evaluate individual intakes for day-to-day patterns and to relate these patterns to the reliability of estimated daily energy and nutrient intake means. Generalized least squares estimators incorporating a simple persistence hypothesis showed the importance of including day-to-day patterns in estimating mean daily intake levels. The simple persistence hypothesis was that day-to-day intakes follow a first-order autoregressive process. Results on the accuracy of mean intake estimates and sample size suggested that for the dietary components examined, greatest gains in accuracy of estimated mean daily intake were generally obtained with the first six days of intake data. This conclusion was, however, highly conditioned by the inclusion of the persistence hypothesis in the calculations. It was concluded that additional investigations exploring alternative and perhaps more physiologically justified persistence hypotheses for food intakes would improve the use of sample data in estimating mean daily intakes of diet components.