Details and implementation of the Box-Cox symmetric class of models and applications for food intake data

• Grant number: 19/02231-6
• Support Opportunities: Regular Research Grants
• Start date: February 01, 2020
• End date: January 31, 2022
• Field of knowledge: Health Sciences - Nutrition - Nutritional Analysis of Population
• Principal Investigator: Giovana Fumes Ghantous
• Grantee: Giovana Fumes Ghantous
• Host Institution: Faculdade de Zootecnia e Engenharia de Alimentos (FZEA). Universidade de São Paulo (USP). Pirassununga , SP, Brazil
• Associated researchers: César Gonçalves de Lima ; José Eduardo Corrente ; Rachel Santos Bueno Carvalho ; Raúl Alejandro Morán Vásquez ; Silvia Lopes de Paula Ferrari
• Associated scholarship(s): 20/03228-6 - Study of computational tools for statistical analysis and preparation of technical reports., BP.TT

Abstract

In nutrition field, it is the great interest of researchers to estimate the usualnutrient intake distribution of a population group based on data obtained from two ormore 24-hour dietary recalls. To estimate this distribution, a Box-Cox transformation is generally used to normalize the data set (or at least tomake the distribution symmetrical) andto apply a back transformation (Dodd et al.; 2006; Tooze et al.; 2010) in order to estimatethe prevalence of nutrient inadequacy. Fumes-Ghantous et al. (2018) proposes an alternativeapproach to estimate the usual nutrient intake distribution through the Box-Coxt model with random intercept(Rigby and Stasinopoulos; 2006), specially contemplatingdata set with high asymmetry and/or with outliers, which is commonly found in nutrientintake data. In addition, Ferrari and Fumes (2017) proposed the Box-Cox symmetric classof distributions, which allows the interpretation of the parameters of the distributions that compose it in terms of quartiles, relative dispersion and skewness, which makes it attractive for regression modeling. The present study proposes the implementation of the Box-Cox symmetric class of distributions as well as the detailing of the functions. The implementation will be done through routine construction in the R program. To test the routines and the models used will be made applications to nutrient intake data of theelderly. (AU)