Abstract
Various methods are available for estimating usual dietary intake distributions. Hence, there is a need for simulation studies to compare them. The methods Iowa State University (ISU), National Cancer Institute (NCI), Multiple Source Method (MSM) and Statistical Program to Assess Dietary Exposure (SPADE) were previously compared in another study, but some results were inconclusive due to the small number of replications used in the simulation. Seeking to overcome this limitation, the present study used 1000 simulated samples for 12 different scenarios to compare the accuracy of estimates yielded by the aforementioned methods. The focus is on scenarios that exhibited the most uncertainty in the conclusions of the mentioned study above, i.e., scenarios with small sample sizes, skewed intake distributions, and large ratios of the between- and within-person variances. Bias was used as a measure of accuracy. For scenarios with small sample sizes (n = 150), the ISU, MSM and SPADE methods generally achieved more accurate estimates than the NCI method, particularly for the 10th and 90th percentiles. The differences between methods became smaller with larger sample sizes (n = 300 and n = 500). With few exceptions, the methods were found to perform similarly.