Mawadah, Nur (2025) Pemodelan Geographically Weighted Log-Logistik 3-Parameter Regression , (Studi Kasus: Prevalensi Stunting di Level Kabupaten/Kota Pulau Sulawesi Tahun 2023). Masters thesis, Institut Teknologi Sepuluh Nopember.
|
Text
6003241012-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy |
Abstract
Geographically Weighted Log-Logistic 3-Parameter Regression (GWLL3R) digunakan untuk menganalisis data prevalensi stunting di 81 kabupaten/kota di Pulau Sulawesi, dengan memanfaatkan dataset tahun 2023 yang diperoleh dari Badan Pangan Nasional (BAPANAS). Mengatasi keterbatasan model global dalam menangani heterogenitas spasial dan data yang miring, model ini menggunakan fungsi pembobot kernel Gaussian tetap (Fixed Gaussian). Estimasi parameter dilakukan menggunakan Metode Maximum Likelihood Estimation (MLE) yang dioptimalkan dengan algoritma Berndt-Hall-Hall-Hausman (BHHH). Hasil analisis menunjukkan bahwa model GWLL3R dengan bandwidth optimum sebesar 13,1043 memberikan kecocokan terbaik, menghasilkan nilai AICc terendah sebesar 582,2952 dibandingkan fungsi kernel lainnya. Selain itu, uji perbandingan model menegaskan bahwa model spasial ini secara signifikan lebih unggul dibandingkan model global LL3R. Hasil analisis mengidentifikasi empat kelompok spasial yang berbeda, dengan persentase kemiskinan dan rasio tenaga kesehatan muncul sebagai faktor penentu paling kuat yang secara signifikan memengaruhi stunting di seluruh lokasi, sementara variabel lain seperti pendidikan hanya menunjukkan efek lokal. Temuan kuantitatif ini menegaskan keunggulan GWLL3R dalam menangkap faktor spesifik lokasi, memberikan dasar yang tepat untuk intervensi kesehatan regional yang terarah.
===================================================================================================================================
Geographically Weighted Log-Logistic 3-Parameter Regression (GWLL3R) to analyze stunting prevalence data across 81 districts/cities in Sulawesi Island, utilizing the 2023 dataset obtained from the National Food Agency (BAPANAS). Addressing the limitations of global models in handling spatial heterogeneity and skewed data, the model employs a Fixed Gaussian kernel weighting function. Parameter estimation was conducted using the Maximum Likelihood Estimation (MLE) optimized with the Berndt-Hall-Hall-Hausman (BHHH) algorithm. The results demonstrate that the GWLL3R model with an optimum bandwidth of 13.1043 achieves the best fit, yielding the lowest AICc value of 582.2952 compared to other kernel functions. Furthermore, model comparison tests confirm that the spatial model significantly outperforms the global LL3R model. The analysis identified four distinct spatial clusters, with poverty percentage and health worker ratio emerging as the most robust determinants significantly affecting stunting across all locations, while other variables such as education showed localized effects. These quantitative findings validate the GWLL3R’s superiority in capturing location-specific factors, providing a precise basis for targeted regional health interventions.
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | AICc Bandwidth, Geografi Kesehatan, GWLL3R, LL3R, Prevalensi Stunting |
| Subjects: | H Social Sciences > HA Statistics > HA29 Theory and method of social science statistics |
| Divisions: | Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49101-(S2) Master Thesis |
| Depositing User: | Nur Mawadah |
| Date Deposited: | 08 Jan 2026 07:08 |
| Last Modified: | 08 Jan 2026 07:08 |
| URI: | http://repository.its.ac.id/id/eprint/129372 |
Actions (login required)
![]() |
View Item |
