Geographically Weighted Multivariate Poisson Regression (Studi Kasus : Pemodelan Jumlah Kematian Ibu, Neonatal Dini, dan Neonatal Lanjut di Provinsi Jawa Tengah)

Triyanto, . (2017) Geographically Weighted Multivariate Poisson Regression (Studi Kasus : Pemodelan Jumlah Kematian Ibu, Neonatal Dini, dan Neonatal Lanjut di Provinsi Jawa Tengah). Doctoral thesis, Institut Teknologi Sepuluh Nopember.

[img]
Preview
Text
1312301004-Disertation.pdf - Published Version

Download (12MB) | Preview

Abstract

Tujuan penelitian ini adalah mengembangkan model multivariate Poisson regression (MPR) dan geographically weighted multivariate Poisson regression (GWMPR), yang mencakup estimasi parameter dan pengujian hipotesis. Pengembangan model dilakukan untuk kovariansi merupakan konstanta dan kovariansi merupakan fungsi variabel bebas. Selanjutnya, model MPR dan GWMPR yang telah dikembangkan diterapkan untuk pemodelan jumlah kematian ibu, neonatal dini, dan neonatal lanjut di Provinsi Jawa Tengah. Hasil penelitian ini, estimasi parameter dilakukan dengan metode maximum likelihood estimation (MLE). Permasalahan dalam estimasi parameter dari model ini, metode MLE tidak dapat menemukan penyelesaian analitis, sehingga diterapkan prosedur iterasi dengan algoritma Newton-Raphson. Pengujian hipotesis dalam model MPR dan GWMPR yang meliputi : uji kesamaan model GWMPR dan MPR, uji parameter serentak, dan uji parameter parsial dilakukan dengan metode likelihood ratio test (LRT). Hasil studi empiris menunjukkan ada perbedaan antara model GWMPR dan MPR. Memperhatikan nilai AIC dan MSE, model GWMPR lebih baik daripada model MPR dalam pemodelan jumlah kematian ibu, neonatal dini, dan neonatal lanjut. Sementara itu, model dengan kovariansi merupakan fungsi variabel bebas lebih baik daripada model dengan kovariansi merupakan konstanta. ================================================================================================================== The aim of this study is to develop a model of multivariate Poisson regression (MPR) and geographically weighted multivariate Poisson regression (GWMPR) which include parameter estimation and hypothesis testing. Development of the models are done for a constant covariance and covariance as a function of the independent variables. Furthermore, the MPR and GWMPR models that have been developed will be applied to modelling the number of mortality of maternal, early neonatal, and late neonatal in Central Java Province. The results of this study, the parameters estimation of the MPR and GWMPR models are done by using maximum likelihood estimation (MLE) method. The problem of parameters estimation for these models, MLE method can not find an analytical solution, so it is applied iterative procedure by the Newton-Raphson algorithm. The hypothesis testing in GWMPR model which include a goodness of fit test, an overall test, and test of individual parameters are done by using Likelihood Ratio Test (LRT) method. The results of empirical studies show the difference between the models of GWMPR and MPR. Considering the value of AIC and MSE, the model of GWMPR is better than the MPR model to modelling of the number of mortality of maternal, early neonatal, and late neonatal. Meanwhile, the model with the covariance as a function of the independent variable is better than a model with a constant covariance.

Item Type: Thesis (Doctoral)
Additional Information: RDSt 519.537 Tri g
Uncontrolled Keywords: geographically weighted, multivariate Poisson regression, data spasial, kematian ibu, neonatal dini, neonatal lanjut.
Subjects: H Social Sciences > HA Statistics
Divisions: Faculty of Mathematics and Science > Statistics > (S3) PhD Theses
Depositing User: Users 13 not found.
Date Deposited: 29 May 2017 02:45
Last Modified: 28 May 2019 02:57
URI: http://repository.its.ac.id/id/eprint/41391

Actions (login required)

View Item View Item