Um estudo sobre o efeito do plano amostral da PNAD Contínua

Abstract

In Brazil, numerous surveys are conducted to estimate various social and economic indicators. Some of these studies are complex and wide ranging, resulting in a large volume of data produced. Therefore, sampling methods are necessary to take into account all the elements that make up the population in a representative sample and thus to construct estimates. With this, the sampling takes place from the collection of data to the analysis of the same. In order to carry out this work we use data from the PNAD-C conducted by the IBGE since 2012. We aim to verify the problems caused by ignoring the sampling plan of the Continuous PNAD in the analysis of data. We will summarize and introduce some more usual sampling plans and methods for analyzing complex sample data. We will also present results of rates that were estimated in order to illustrate the purpose of the work. Stata 15 software was used to calculate estimates, confidence intervals, standard errors and EPAs, considering and not considering the sample plan of the PNAD-C. We observed that when we ignored the sampling plan, there was a difference in the values obtained for the estimates of the means if we compare with the estimated means obtained under the sampling plan of the research and also the underestimation of the errors occurred that directly affected the confidence intervals produced.