Analysis of Multidimensional Stunting Intervention Factor Using Mixed Model

Abstract

Abstract The mixed model combines fixed effect for all groups and random effect representing the diversity inter groups in the model (province) to increase the model precision. This study provides information on the significance of multidimensional stunting intervention factors (predictor variables) on stunting prevalence (response variables as indicator 2.2.1 Sustainable Development Goals/SDGs) with district/city as observation units. Using official data from Statistics Indonesia (National Socio Economic Survey) and Ministry of Health (Basic Health Research), this study expects to be one basis of information for the government, stakeholders, and further research to accelerate Indonesia’s SDGs targets in 2030. Comparison of classical linear mixed model method and linear mixed model with Least Absolute Shrinkage and Selection Operator (Lasso) variable selection conduct with relatively better results of mixed linear modelling with Lasso. The results showed that the predictor variables, namely complete immunization, ease of access to health facilities, diversity of food intake, improve water, food expenditure per capita, children’s participation in early childhood education, maternal education, and ownership of National Health Insurance for toddlers, significantly affected the stunting prevalence decrease. The predictor variables, namely low birth weight, households with social protection cards, and the percentage of poor people, significantly increase the stunting prevalence.