Application of Logistic Regression Modeling for Complex Survey Data on Education Continuity of Poor Households Children

Abstract

Many population-based surveys such as the National Socio-Economic Survey (Susenas) are built with complex sampling assumptions, namely probabilistic, stratified, and multistage sampling, with unequal weights for each observation. This complex design must be taken into account in order to have reliable results when doing modeling. The model that is often used when using survey data is logistic regression. The purpose of this study is to determine a logistic regression model with a complex sample design, and to show how it is estimated using a package survey from the R software. As an illustration, the 2019 Susenas data of East Java Province will be used as an application to correct the influence of the sample design in estimating risk factors related to the chances of children 7-18 years old in poor households continuing their education. The results show that the variables of gender and mother's education significantly affect the continuity of the education of children 7-18 years old in poor households.