Sandy Burden1, Yasmine Probst, David Steel, Linda Tapsell. 1. Centre for Statistical and Survey Methodology, University of Wollongong, Northfields Avenue, Wollongong, New South Wales 2522, Australia. sburden@uow.edu.au
Abstract
OBJECTIVE: To assess the impact of the complex survey design used in the 2007 Australian National Children's Nutrition and Physical Activity Survey (ANCNPAS07) on prevalence estimates for intakes of groups of foods in the population of children. DESIGN: The impacts on prevalence estimates were determined by calculating design effects for values for food group consumption. The implications of ignoring elements of the sample design including stratification, clustering and weighting are discussed. SETTING: The ANCNPAS07 used a complex sample design involving stratification, a high degree of clustering and estimation weights. SUBJECTS: Australian children aged 2-16 years. RESULTS: Design effects ranging from <1 to 5 were found for the values of mean consumption and proportion of the population consuming the food groups. When survey weights were ignored, prevalence estimates were also biased. CONCLUSIONS: Ignoring the complex survey design used in the ANCNPAS07 could result in underestimating the width of confidence intervals, higher mean square errors and biased estimators. The magnitude of these effects depends on both the parameter under consideration and the chosen estimator.
OBJECTIVE: To assess the impact of the complex survey design used in the 2007 Australian National Children's Nutrition and Physical Activity Survey (ANCNPAS07) on prevalence estimates for intakes of groups of foods in the population of children. DESIGN: The impacts on prevalence estimates were determined by calculating design effects for values for food group consumption. The implications of ignoring elements of the sample design including stratification, clustering and weighting are discussed. SETTING: The ANCNPAS07 used a complex sample design involving stratification, a high degree of clustering and estimation weights. SUBJECTS: Australian children aged 2-16 years. RESULTS: Design effects ranging from <1 to 5 were found for the values of mean consumption and proportion of the population consuming the food groups. When survey weights were ignored, prevalence estimates were also biased. CONCLUSIONS: Ignoring the complex survey design used in the ANCNPAS07 could result in underestimating the width of confidence intervals, higher mean square errors and biased estimators. The magnitude of these effects depends on both the parameter under consideration and the chosen estimator.
Authors: Yue Wu; Wei Zhang; Lin Zhang; David C Schwebel; Peishan Ning; Xunjie Cheng; Xin Deng; Li Li; Jing Deng; Guoqing Hu Journal: Eur J Public Health Date: 2017-04-01 Impact factor: 3.367