Using the overdispersed exponential family to estimate the distribution of usual daily intakes of people aged between 18 and 28 in Taiwan.

Assessment of usual dietary intake of a population is essential to understand the nutrition status of the population. Methods like 24-hour dietary recall and food records are commonly used for this task. However, within-individual variation exists in the food record data. Methods considering measurement errors have been used to evaluate the association between nutrient intakes and diseases. Few have been devoted to estimating the distribution of usual daily intakes. This paper proposes applying the formulation of the overdispersed exponential family to estimate the distribution of usual nutrient intake and applies the adjustments developed by Liu to reduce the variance of the distribution. The proposed method has the advantages of working on the original scale of data and implementation convenience. The adjustment of the variance is carried out by dividing the variance into within-individual variance and among-individual variance. The adjusted variance, then, is used to estimate the distribution of the usual daily intake. Sampling weights are considered in all the steps, except the estimation of the ratio of within to among variance. The data for this study are from the Nutritional and Health Survey in Taiwan conducted between 1993 and 1996 (NAHSIT, 1993-1996). An independent external set of data for people aged between 18 and 28 years (49 males and 20 females) is used to estimate the ratio of within to among individual variance. Due to the availability of data in estimating the ratio of within to among individual variance, 430 males and 422 females in the NAHSIT sample are used to estimate the distribution of usual daily intakes for people aged between 18 and 28 in Taiwan. The distribution also allows us to estimate the proportion of people who do not meet the recommended daily nutrient allowance (RDNA). The results show that the intakes of calcium and vitamin E of this group of people fall far below RDNA. Copyright 2001 John Wiley Sons, Ltd.