Análise de correspondência para avaliação de indicadores socioeconômicos, demográficos e de mortalidade

Abstract

We present a study of the associations, similarities or dissimilarities among selected socioeconomic, demographic and mortality indicators for all states and regions in Brazil, through the adoption of a multivariate statistical technique called Correspondence Analysis (CA). The indicators were gathered from IDB-2009 (Indicators and Basic Data- 2009) and also from IPEA (Applied Economic Research Institute). Some descriptive statistics for all indicators used in the analysis are presented, along with corresponding box-plots. As CA was strictly developed for categorical data, and the indicators we selected were all numerical variables, two grouping methods were used in order to categorize such variables: k- means and percentiles. The analysis of the categorized variables through the k-means method presented a solution which best explained the overall variance in the CA solution and was therefore adopted. However, the results obtained through the percentiles method presented similar results. The scaling correction known as “doubling” was then applied to the k-means generated data and another CA solution was obtained. These results were similar to the previous one, but the visual interpretation was more straightforward. We could thus analyze the relative position of states, their similarities in terms of the indicators adopted and their association with such indicators. AC was selected for its unique feature of presenting the solution in a clear, easy-to-interpret graphic form. All calculations were made in the open- source programming code R.