CD Skripsi
Perbandingan Fungsi Pembobot Kernel Pada Geographically Weighted Negative Binomial Regression Dalam Memodelkan Kematian Bayi Indonesia
Infant Mortality Rate (IMR) in Indonesia remains a significant challenge in achieving the Sustainable Development Goals (SDGs), particularly target 3.2 related to reducing newborn and under-five mortality. Unequal access to healthcare services and socioeconomic conditions across different regions in Indonesia result in varying IMR, necessitating spatial regression analysis to model and identify spatially contributing risk factors. This study aims to compare four adaptive kernel weighting functions, namely Gaussian, bisquare, tricube, and exponential, in modeling infant mortality in Indonesia in 2022 using Geographically Weighted Negative Binomial Regression (GWNBR) and to identify influencing factors. The analysis was conducted using 2022 infant mortality data as the response variable and seven predictor variables: low birth weight (X1), unattended births (X2), antenatal care visits (X3), vulnerable maternal age (X4), smoking mothers (X5), poverty (X6), and inadequate access to safe drinking water (X7). The results show that the adaptive exponential kernel produces the best model with the lowest AIC and RMSE values. Risk factors such as low birth weight, unattended births, antenatal care visits, vulnerable maternal age, poverty, and inadequate access to safe drinking water significantly influence the spatial distribution of infant mortality.
Keywords: Infant mortality, Geographically Weighted Negative Binomial Regression, kernel weighting function, Akaike Information Criterion, Root Mean Squared Error.
Tidak tersedia versi lain