Impactos do desenho amostral complexo na análise de dados da PNAD

Abstract

All kinds of researches are performed in Brazil and in the world currently. The amount of data available is huge. Sampling methods are very relevant in this context and, therefore, they were the subject of this study. In this sense, it is essential to take precautions throughout the sampling process, from before collecting data until the analysis phase. In this work we adopt the National Household Sample Survey (PNAD) conducted by the Brazilian Institute of Geography and Statistics (IBGE) as motivation for the application of the studied methods. The main goal of this work is to evaluate the potential effects of the lack of consideration of sample design features in the data analysis of PNAD. We also present a summary of the main types of sampling designs and inference methods for complex sampling in order to introduce the subject. The statistical software Stata has helped in the estimation of the mean and its standard error of the variable "monthly household income per capita" taking sample design into account or not. In the case of PNAD, some of the consequences observed when the design was ignored were: (i) underestimation of standard errors and, therefore, biased confidence intervals; and (ii) different estimated means compared to those found when sampling design was considered.