APPLICATION OF ROBUST GEOGRAPHICALLY WEIGHTED REGRESSION (RGWR) FOR MODELING STUNTING TODDLERS IN WEST JAVA 2024

RIZKIA SYEFIRA, BANISA (2026) APPLICATION OF ROBUST GEOGRAPHICALLY WEIGHTED REGRESSION (RGWR) FOR MODELING STUNTING TODDLERS IN WEST JAVA 2024. FAKULTAS MATEMATIKA DAN ILMU PENGETAHUAN ALAM, UNIVERSITAS LAMPUNG.

[img]
Preview
 Text
ABSTRAK.pdf
Download (16kB) | Preview
[img] Text
SKRIPSI FULL.pdf
Restricted to Repository staff only
Download (3MB) | Request a copy
[img]
Preview
 Text
SKRIPSI TANPA BAB PEMBAHASAN.pdf
Download (2MB) | Preview

Abstract

Regresi linear global mengasumsikan bahwa hubungan antara variabel respon dan variabel prediktor bersifat sama pada seluruh wilayah sehingga kurang mampu mengakomodasi heterogenitas spasial. Geographically Weighted Regression (GWR) dikembangkan untuk mengatasi permasalahan tersebut, namun metode ini sensitif terhadap keberadaan pencilan yang dapat menyebabkan estimasi parameter menjadi tidak stabil. Salah satu cara untuk mengatasi pencilan pada GWR adalah dengan menggunakan metode penduga Robust yaitu Estimasi-MM. Penelitian ini bertujuan menerapkan Robust Geographically Weighted Regression (RGWR) dengan Estimasi-MM untuk memodelkan jumlah balita stunting di 27 kabupaten/kota di Jawa Barat tahun 2024. Variabel prediktor yang digunakan meliputi jumlah fasilitas kesehatan, persentase ASI eksklusif, persentase rumah tangga dengan sanitasi layak, tingkat melek huruf, dan jumlah balita dengan gizi buruk. Hasil analisis menunjukkan adanya heterogenitas spasial dan 9 lokasi pencilan. Model RGWR yang dibangun menggunakan fungsi pembobot kernel Bisquare adaptif menghasilkan kinerja yang lebih baik dibandingkan model GWR, dengan nilai AIC sebesar 404,0302 dan MSE sebesar 0,0584. Selain itu, pengaruh variabel prediktor terhadap jumlah balita stunting berbeda pada setiap wilayah, yang menunjukkan adanya variasi karakteristik antar kabupaten/kota di Jawa Barat. Kata Kunci: Robust Geographically Weighted Regression, Heterogenitas Spasial, Estimasi-MM, Pencilan, Balita Stunting. Global linear regression assumes that the relationship between the response and predictor variables is constant across all regions, making it less capable of accommodating spatial heterogeneity. Geographically Weighted Regression (GWR) was developed to address this issue. However, the method is sensitive to outliers, which may lead to unstable parameter estimates. One way to overcome outliers in GWR is to use Robust techniques, one of which is the MM-Estimator. This study aims to apply Robust Geographically Weighted Regression (RGWR) with MM-estimation to model the number of stunting toddlers in 27 regencies and cities in West Java in 2024. The predictor variables include the number of health facilities, the percentage of exclusive breastfeeding, the percentage of households with proper sanitation, the adult literacy rate, and the number of malnourished toddlers. The results indicate the presence of spatial heterogeneity and nine outlier locations. The RGWR model, constructed using an adaptive Bisquare kernel weighting function, outperformed the GWR model with an AIC value of 404.0302 and an MSE value of 0.0584. Furthermore, the effects of predictor variables on the number of stunting toddlers varied across regions, indicating differences in regional characteristics among regencies and cities in West Java. Keywords: Robust Geographically Weighted Regression, Spatial Heterogeneity, MM-estimation, Outlier, Stunting Toddlers.

Item Type: Other
Subjects: ?? 510 ??
Divisions: Fakultas MIPA > Prodi Matematika
Depositing User: 2606707619 Digilib
Date Deposited: 09 Jul 2026 02:43
Last Modified: 09 Jul 2026 02:43
URI: http://digilib.unila.ac.id/id/eprint/102249

Actions (login required)

View Item View Item