Models with measurement errors: application a nutritional data

• Grant number: 12/17454-1
• Support Opportunities: Regular Research Grants
• Start date: August 01, 2013
• End date: July 31, 2015
• Field of knowledge: Health Sciences - Nutrition - Nutritional Analysis of Population
• Principal Investigator: José Eduardo Corrente
• Grantee: José Eduardo Corrente
• Host Institution: Instituto de Biociências (IBB). Universidade Estadual Paulista (UNESP). Campus de Botucatu. Botucatu , SP, Brazil

Abstract

The analysis of nutritional data has been a major challenge for both professionals in the field of nutrition as the area of Statistics. Several methods of analysis of consumption data have been proposed, such as Iowa State University (ISU) method, Iowa State University for Food (ISUF) method and National Cancer Institute (NCI) using mixed models with measurement errors in order to remove the within-person variability, the main effect observed due to the variability of the diet. Such models are needed to estimate the prevalence of inadequate nutrient consumption that is made using the Dietary Reference Intakes (DRI) as cutoffs. The main problem lies in the fact that various nutrients have strongly asymmetric distribution. In these cases, the use of Box-Cox transformations are not effective to correctly estimate the consumption distribution. Such transformations are used in both the ISU method as in NCI, but the adjustments are not adequate and the prevalence of consumption are under or overestimated. The objective of this study is to propose a new approach for estimating the distribution of nutrient intake in a population for asymmetric data through the use of the program GAMLSS (Generalized Aditive models for Location, Scale and Shape) and apply them to data obtained nutritional surveys of adults and older population in the cities of São Paulo, Botucatu and Avare, SP, and compared with the NCI and ISU methods as well as the estimates of the prevalence of inadequate intake. (AU)