Comparison of methods for the treatment of anthropometric measures of POF 2008-2009

Abstract

The 2008-2009 Pesquisa de Orçamentos Familiares (Household Budget Survey) is a nationwide sample survey that collected anthropometric data on height and weight of individuals in Brazil. Due to the size of the research, the collection process allows contamination in the collected data by non-sampling errors and non-response. Such errors can affect the indicators of malnutrition, prevalence of overweight persons and obesity, and produce differing effects in different population segments. In this study, the Cidaq approach – the methodology employed in POF
2008-2009 to tackle these problems and preserve the quality of the data – was compared with two other approaches – namely, the TRC algorithm and the Bacon algorithm, both coupled with the Poem algorithm. Such comparisons are essential to ensure the choice of the best method future research efforts in order to ensure reliability of the data used in population studies that support the planning of public policies in health, nutrition, social assistance and others. The approaches were compared by simulation through impact in mean, standard deviation and correlation estimates of the weight and height. The Cidaq approach proved to be more efficient than the others under the parametric simulation, while the Bacon approach showed to be better under the non parametric simulation.

DOI http://dx.doi.org/10.20947/S0102-309820160008