Within-person variability of nutrient intake in a group of Hawaiian men of Japanese ancestry.

Within-person variability of nutrient intake is examined using data on 7 consecutive days for 329 men. The variability of nutrient intake within persons differs for the nutrients examined. Some nutrients, such as starch, have small within-person components of variance and are therefore highly reliable even using a single 24-h recall. Others, such a polyunsaturated fatty acids, have large within-person components of variance and are thus less reliable. Within-person variability affects the power of a study to detect differences in nutrient intake between groups and also affects the magnitude of calculated correlation coefficients. For the least reliable nutrient it is estimated that the correlation would be attenuated to two-thirds its true size. Eating patterns in the cohort appear to be nonrandom. Using this nonrandomness of nutrient intake it may be possible to reduce the within-person component of variance using less recalls than is suggested if random eating patterns are assumed.