Combinação de dados amostrais de duas edições da Pnad contínua

Abstract

Survey estimates are often affected by sampling and non-sampling errors. Therefore, combining data from more than one survey can be beneficial for producing more consistent estimators for the population of interest, as it expands the total sample size, reducing sampling errors. You can combine not only separate surveys, but also continuous samples from the same survey, as well as overlapping panel data into a panel survey. In this context, the objective of this study was to combine data from the the Pesquisa Nacional por Amostra de Domicílios Contínua (PNADC) collected in 2021 and 2022, with the aim of reducing sampling errors in the estimates of the brazilian population. To this end, the method developed by Dzikiti, Vieira and Girdler-Brown (2023) was used to seek combined estimators from two or more surveys, using as weights the inverse of the coefficient of variation and the inverse of the effect of the sampling plan enlarged. We also used the inverse of the sample size and the inverse of the variance proposed by Fox (2010) in her studies on meta-analysis. To apply the studied methods, the quantitative variable “rendimento efetivo no trabalho principal” (VD4020) was used. It is concluded that, when combining the surveys, there was an improvement in the accuracy of the estimates with standard errors smaller than what would be observed when only one survey is used.