Pemodelan Regresi Hurdle Negative Binomial dengan Variabel Dependen Tersensor Kanan Pada Kasus Tetanus Neonatorum di Indonesia

Rusdiana, Riza Yuli (2017) Pemodelan Regresi Hurdle Negative Binomial dengan Variabel Dependen Tersensor Kanan Pada Kasus Tetanus Neonatorum di Indonesia. Masters thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 1315201013-Master-Theses.pdf]
Preview
Text
1315201013-Master-Theses.pdf - Published Version

Download (2MB) | Preview

Abstract

Model regresi Hurdle Negative Binomial (HNB) adalah metode yang dapat digunakan untuk variabel dependen bertipe data count dengan banyak observasi yang bernilai nol (excess zero) dan terjadi overdispersion. Model HNB menggunakan pendekatan dua bagian (two part model), yaitu bagian pertama untuk mengestimasi variabel dependen bernilai nol dan bagian kedua mengestimasi variabel dependen yang bernilai bulat non-negatif. Untuk kasus tertentu variabel dependen tersensor pada nilai tertentu. Jenis sensor yang akan digunakan yaitu sensor kanan. Penelitian ini akan melakukan kajian teori, kajian simulasi dan kajian terapan pada model regresi Censored Hurdle Negative Binomial (CHNB). Pada kajian teori dilakukan estimasi parameter model regresi CHNB menggunakan metode maksimum likelihood menghasilkan persaman tidak closed form, sehingga untuk menyelesaikan estimasi parameter digunakan metode iterasi Newton Rapshon. Berdasarkan hasil simulasi semakin besar data mengalami penyensoran maka semakin besar pula performa model regresi CHNB dan semakin besar ukuran sampel semakin besar performa model regresi CHNB. Di sisi lain, adapun pemodelan regresi CHNB terhadap kasus tetanus neonatorum di Indonesia didapatkan kedua model yaitu zero hurdle model dan truncated negative binomial model. Variabel imunisasi TT2+, imunisasi TT5 dan persalinan di fasilitas kesehatan berpengaruh terhadap jumlah kasus tetanus neonatorum.

=========================================================================================

Hurdle Negative Binomial (HNB) regression model is a method which can be used for dependent variable of count data type with many zeros and overdispersion condition. The HNB model uses a two-part approach (two part model) i.e. the first part for zero count and another part for positive count.The dependent variable in such cases is censored for some values. The right censored is used in this research. Censored Hurdle Negative Binomial (CHNB) regression model is applied on the theory, simulation and empirical studies.The results of theoretical studiesindicate that the equations to obtain estimated parameters are not closed form, then a numerical method with Newton Raphson iteration is used. Based on the result of the simulation, the larger the censored data and the larger of sample size give the better performance CHNB regression model. On the other hand, the result of empirical studies for tetanus neonatorum case in Indonesia is obtained both hurdle model and truncated negative binomial model. Variable of TT2+ immunization, TT5 immunization, and labor in health facility affected number of tetanus neonatorum case.

Item Type: Thesis (Masters)
Uncontrolled Keywords: Hurdle Negative Binomial; Tersensor Kanan; Tetanus Neonatorum; Two Part Model
Subjects: Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression
Divisions: Faculty of Mathematics and Science > Statistics > 49101-(S2) Master Thesis
Depositing User: - RIZA YULI RUSDIANA
Date Deposited: 06 Apr 2017 02:42
Last Modified: 06 Mar 2019 04:37
URI: http://repository.its.ac.id/id/eprint/3066

Actions (login required)

View Item View Item