Characterizing the Distribution of Macronutrient Intake among U.S. Adults: A Quantile Regression Approach

Since the risk of dietary inadequacy or excess is greater at the tails of the nutrient intake distributions than at the mean, marginal effects of explanatory variables estimated at the conditional mean using ordinary least squares may be of limited value in characterizing these distributions. Quantile regression is effective in this situation since it can estimate conditional functions at any part of the distribution. Quantile regression results suggest that age, education, and income have a larger influence at intake levels where the risk of excess is greater compared with intake levels where the risk of excess is lower. Copyright 2002, Oxford University Press.