This study applied geo-additive regression modelling on high-dimensional data with metrical and categorical predictors, using data from the latest Nigeria Demographic and Health Survey (NDHS-6). The sampled data comprises ninety-three predictors of total children ever born by eighty women of age group between fifteen and forty-nine years. Three penalty-based variable selection criteria—Elastic Net, Smoothly Clipped Absolute Deviation (SCAD), and Minimax Concave Penalty (MCP) were employed to reduce the dimensionality of the data. Geo-additive regression model were thereafter applied on the selected predictors to determine the impact of metrical, categorical and spatial predictors on the response variable. Findings revealed that predictors such as contraceptive use, age at first birth, and marital status are major determinants of total children ever born by a woman in Nigeria. Furthermore, spatial analysis revealed regional disparities in fertility rates within Nigeria, with notably higher fertility rate in northeastern states. This study’s findings have broad applications across disciplines. By providing robust methodologies for handling complex datasets, this research supports evidence-based decision-making in public health, agriculture, and environmental policy. Ultimately, these findings contribute to efforts aimed at promoting sustainable development and enhancing maternal and child health outcomes in Nigeria and similar contexts globally.
Read full abstract