Premium
Predictive Validity of Dietary Assessment Questionnaires
Author(s) -
Arab Lenore,
Tseng Chihong,
Cambou Mary Catherine
Publication year - 2010
Publication title -
the faseb journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.709
H-Index - 277
eISSN - 1530-6860
pISSN - 0892-6638
DOI - 10.1096/fasebj.24.1_supplement.96.2
Subject(s) - akaike information criterion , statistics , covariate , bayesian information criterion , mathematics , correlation , criterion validity , cross validation , deviance information criterion , regression analysis , population , bayesian probability , econometrics , medicine , construct validity , environmental health , bayesian inference , psychometrics , geometry
Traditionally correlation analyses and underreporting have been used for determining the value of a dietary assessment tool. We explored the use of predictive validation using log regression models with and without covariates and interaction terms to compare 6 days of 24‐hour recalls with food frequency questionnaires (DHQ). The data derives from a population of African‐Americans and Caucasians from Los Angeles participating in the Energetics Study, designed to test the validity of a web‐based automated 24‐hour recall (DietDay) using doubly labeled water as a marker of true energy intake. We observed that underreporting was considerably lower, with the recalls capturing 84% of energy expenditure as compared with 73% by the DHQ, while the correlation analyses showed similar correlations for both dietary assessment tools. We used Akaike's Information Criterion (AIC) to compare models and Bayesian Information Criterion (BIC) as a form of regularization to prevent over fitting. With the backward model selection procedure the adjusted R‐square of the final model was 0.511 for DietDay with covariates, compared to 0.103 without covariates. The Adjusted R‐squares for DHQ was 0.509 with and 0.118 without covariates. This quantitative approach provides more robust comparisons of the predictive validity of dietary assessment tools. Supported by NIH R01CA105048.