[Brazilian Oral Health Survey (SB Brazil 2003): data do not allow for population estimates, but correction is possible].

The Brazilian Oral Health Survey (SB Brazil 2003) was the most comprehensive study on oral health conditions ever conducted in Brazil. Probabilistic sampling methods were applied in order for the collected data to represent the population age groups selected in the 5 regions of the country. However, this was not possible because the sampling process was never concluded, which would require estimation of the sample weights and identification of the sample's other structural variables (selection strata and primary sampling units). This paper describes the SB Brazil 2003 sample design, formulates the inclusion probabilities in the multiple selection stages, and proposes strategies for estimating the sample weights. The strategy to define the sample weights and identify the sample's structural variables hinges on retrieving data that should have been recorded in the reports produced during the survey, and in their absence, in information available from the Brazilian Institute of Geography and Statistics (IBGE) and Ministry of Education and Culture (MEC), as valid proxies.