Modelos inflacionados de zeros: aplicações na análise de um questionário de frequencia alimentar

This study aimed at using zero-inflated models to analyze a food frequency questionnaire (F.F.Q.) applied to elderly individuals from a medium-sized city in inner São P aulo state, investigating the models that can be best fitted to the data according to the percent of zeros found and the factors in...

ver descrição completa

Detalhes bibliográficos
Autores: Fumes, Giovana [UNESP], Corrente, José Eduardo [UNESP]
Tipo de documento: artigo
Estado:Versão publicada
Data de publicação:2010
País:Brasil
Recursos:Universidade Estadual Paulista (UNESP)
Repositório:Repositório Institucional da UNESP
Idioma:português
OAI Identifier:oai:repositorio.unesp.br:11449/140334
Acesso em linha:http://jaguar.fcav.unesp.br/RME/fasciculos/v28/v28_n1/indice_v28_n1.php
http://hdl.handle.net/11449/140334
Access Level:Acceso aberto
Palavra-chave:Negative binomial distribution
Poisson distribution
Zero-inflated distribution
Food frequency questionnaire
Overdispersion
Distribuição binomial negativa
Distribuição de Poisson
|Distribuição inflacionada de zero
Questionário de frequência alimentar
Superdispersão
Descrição
Resumo:This study aimed at using zero-inflated models to analyze a food frequency questionnaire (F.F.Q.) applied to elderly individuals from a medium-sized city in inner São P aulo state, investigating the models that can be best fitted to the data according to the percent of zeros found and the factors influencing both intake and non-intake of certain food types. The response variable used was the monthly intake frequency of foods from the F.F.Q., and the associated co-variates were related to socio-demographic characteristics concerning prevention and reported morbidities. The Poisson, Negative Binomial, Zero-Inflated Poisson and Zero-Inflated Negati e Binomial models were used. For a data set with up to 10% of zeros, the usual Negative Binomial model showed to be adequate. Due to data amplitude, the inflated models show ed to b e adequate only at the p ercentage of 1 0 % to 5 0 % of zeros. It was also shown that the probability of consumption of the selected food types was associated with gender, nutritional status, medication use and physical activity practice. The probability of non-consumption was observed in association with gender, bow el functioning and daily water consumption .