Estimation of the distribution of protein requirements in a population

Craig Alan Johnson, Purdue University

Abstract

Nitrogen balance techniques have been the premier method for estimating protein requirements in humans for over 125 years. It has been suggested that nitrogen balance methods should no longer be considered the gold standard for estimating protein requirements in a population. However, no technique has been accepted as a suitable replacement. Improvements in the methods of analyzing nitrogen balance data could enhance its utility. A model for the distribution of the protein requirement in a population was developed, which incorporates sources of variation in nitrogen balance experiments. A methodology for parameterizing these models using laboratory data was established. Applying this methodology to data on the protein requirements of young and elderly women produced an estimate of the true distribution of protein requirements of this population. This method can be used in any application of inverse prediction based on simple linear regression with a small number of observations. The impact of the sources of variation on the distribution of protein requirements was assessed. The form of the distribution of protein requirements in a population is not known. Recent studies suggest that this distribution is lognormal. It was shown that it is possible for the requirement distribution observed in a laboratory to appear lognormal when the true requirement distribution is normal. A new procedure for analyzing nitrogen balance data was proposed that generates reasonable estimates of the mean (or median) and the 97.5 th percentile of the true requirement distribution.

Degree

Ph.D.

Advisors

McCabe, Purdue University.

Subject Area

Statistics|Nutrition

COinS