Model Regresi Mixed Geographically Weighted Bivariate Poisson Generalized Inverse Gaussian (Studi Kasus: Jumlah Kematian Ibu dan Post Neonatal di Jawa Timur Tahun 2023)

Rizkita, Maharani Endah (2025) Model Regresi Mixed Geographically Weighted Bivariate Poisson Generalized Inverse Gaussian (Studi Kasus: Jumlah Kematian Ibu dan Post Neonatal di Jawa Timur Tahun 2023). Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6003231021-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2027. Download (12MB) | Request a copy

Abstract

Model Mixed Geographically Weighted Bivariate Poisson Generalized Inverse Gaussian Regression (MGWBPGIGR) merupakan pengembangan model Bivariate Poisson Generalized Inverse Gaussian Regression (BPGIGR) dan Geographically Weighted Bivariate Poisson Generalized Inverse Gaussian Regression (GWBPGIGR). Model tersebut dibangun untuk mengatasi kasus overdispersi dengan data cacahan yang berkorelasi dan dipengaruhi oleh variabel global yang secara signifikan mempengaruhi semua lokasi pengamatan dan variabel lokal yang secara signifikan mempengaruhi lokasi tertentu. Kajian teori dilakukan untuk mendapatkan penaksir parameter model GWBPGIGR dan MGWBPGIGR menggunakan metode Maximum Likelihood Estimation (MLE) dengan iterasi numerik Berndt-Hall-Hall-Hausman (BHHH). Selanjutnya, statistik uji untuk pengujian hipotesis parameter model GWBPGIGR dan MGWBPGIGR ditentukan menggunakan metode Maximum Likelihood Ratio Test (MLRT). Kajian aplikasi dilakukan dengan mengaplikasikan model GWBPGIGR dan MGWBPGIGR pada data jumlah kematian ibu dan post neonatal di Jawa Timur Tahun 2023 dengan menggunakan lima variabel prediktor dan dua variabel exposure, serta pembobot spasial yang digunakan yaitu fungsi kernel adaptive Gaussian. Dari penelitian ini diperoleh 38 model MGWBPGIGR untuk jumlah kematian ibu dan post neonatal. Model MGWBPGIGR menghasilkan pengelompokan lokasi atau wilayah berdasarkan variabel yang signifikan hanya ditemukan terdapat perbedaan pada jumlah kematian post neonatal dimana terdapat dua pengelompokan kabupaten/kota di Jawa Timur. Sedangkan, jumlah kematian ibu di semua kabupaten/kota dipengaruhi oleh prediktor yang sama. Variabel persentase rumah tangga yang memiliki sanitasi layak merupakan variabel yang berpengaruh secara global di seluruh model MGWBPGIGR baik untuk jumlah kematian ibu dan post neonatal serta terdapat beberapa variabel lain yang hanya berpengaruh pada beberapa model MGWBPGIGR, dimana variabel-variabel tersebut dinamakan variabel yang berpengaruh secara lokal. Hasil penelitian ini juga menunjukkan bahwa model MGWBPGIGR memiliki performa yang lebih baik jika dibandingkan dengan model GWBPGIGR dan model BPGIGR.
==================================================================================================================================
The Mixed Geographically Weighted Bivariate Poisson Generalized Inverse Gaussian Regression (MGWBPGIGR) model is a development of the. Bivariate Poisson Generalized Inverse Gaussian Regression (BPGIGR) model and the Geographically Weighted Bivariate Poisson Generalized Inverse Gaussian Regression (GWBPGIGR) model. These models are built to overcome overdispersion cases with correlated count data and are influenced by global variables that significantly affect all location observations and local variables that significantly affect specific-location observations. Theoretical review was done to obtain parameter estimate of the GWBPGIGR and MGWBPGIGR models by the Maximum Likelihood Estimation (MLE) with the Berndt-Hall-Hall-Hausman (BHHH) iteration method. Meanwhile, the statistical test for hypothesis testing of the GWBPGIGR and MGWBPGIGR models were determined by using the Maximum Likelihood Rasio Tests (MLRT) method. The GWBPGIGR and MGWBPGIGR models will be applied to model the number of maternal mortality and post neonatal mortality in East Java 2023 using five predictor variables and two exposure variables, and the spatial weights used were the adaptive Gaussian kernel. From this study obtained 38 MGWBPGIGR models for the number of maternal mortality and the number of post neonatal mortality. The MGWBPGIGR model resulted in the grouping of the locations or regions based on significant variables, only found differences in the number of post neonatal mortality where there are two groups of districts/cities in East Jawa. Meanwhile, the number of maternal mortality in all districts/cities was influenced by the same predictors. Percentage of household with improved sanitation is a globally influential variable in all MGWBPGIGR models for both the number of maternal mortality and post neonatal mortality, and there are several other variables that influence only some MGWBPGIGR models, where these variables are called locally influential variables. The result also show that the MGWBPGIGR model has better performance compared to GWBPGIGR and BPGIGR models.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	BHHH, MGWBPGIGR, MLE, MLRT, Overdispersi, Overdispersion
Subjects:	Q Science > QA Mathematics > QA276 Mathematical statistics. Time-series analysis. Failure time data analysis. Survival analysis (Biometry) Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression Q Science > QA Mathematics > QA401 Mathematical models.
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49101-(S2) Master Thesis
Depositing User:	Maharani Endah Rizkita
Date Deposited:	03 Feb 2025 08:10
Last Modified:	03 Feb 2025 08:10
URI:	http://repository.its.ac.id/id/eprint/117286

Actions (login required)

View Item