Use of regression adjustments for ascertaining nutrient intakes.

Occasionally, in nutritional studies using human subjects, a representative subsample is taken from the entire study sample, and nutrient intake data are ascertained by two methods, with one method assumed to be more accurate than the other. For logistic reasons, intake data for the remaining subjects in the study sample are ascertained by the less accurate method only. The objective is to predict the intake data for the more accurate method in the entire study sample on the basis of the statistical information obtained in the subsample. This article presents a regression-adjustment method for accomplishing the foregoing objective.