Pemodelan Generalized Poisson Regression (GPR) Dan Negative Binomial Regression (NBR) Untuk Mengatasi Overdispersi Pada Jumlah Kematian Bayi Di Kabupaten Probolinggo

Chaniago, Amara Deviana (2022) Pemodelan Generalized Poisson Regression (GPR) Dan Negative Binomial Regression (NBR) Untuk Mengatasi Overdispersi Pada Jumlah Kematian Bayi Di Kabupaten Probolinggo. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 10611810000026-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (4MB)

Abstract

Data Dinas Kesehatan Kabupaten Probolinggo tahun 2020 menyatakan bahwa Angka Kematian Bayi (AKB) di Kabupaten Probolinggo sebesar 8,11. Angka tersebut tergolong tinggi jika dibandingkan dengan AKB Provinsi Jawa Timur sebesar 6,3. Sehingga, perlu dilakukan analisis untuk mengetahui faktor-faktor yang memengaruhi jumlah kematian bayi di Kabupaten Probolinggo tahun 2020. Regresi Poisson merupakan analisis yang sesuai dalam pemodelan kasus data diskrit. Namun, regresi Poisson mensyaratkan kondisi equidispersi yang sulit dipenuhi. Pada umumnya, sering ditemui data diskrit dengan nilai varian yang lebih besar daripada mean yang disebut dengan overdispersi. Metode yang dapat digunakan untuk mengatasi kondisi overdispersi di antaranya adalah metode Generalized Poisson Regression (GPR) dan Negative Binomial Regression (NBR). Objek penelitian terdiri dari variabel respons, yaitu jumlah kematian bayi, serta variabel prediktor yang diduga memengaruhi jumlah kematian bayi, di antaranya faktor kesehatan ibu hamil dan bayinya, faktor kebersihan dan gizi, serta faktor peningkatan imunitas bayi yang terdiri dari 9 variabel, serta 2 variabel tambahan yang diperoleh melalui publikasi BPS Kabupaten Probolinggo, yaitu jumlah tenaga kesehatan dan jumlah fasilitas kesehatan. Hasil analisis diperoleh bahwa terjadi kasus overdispersi sehingga analisis GPR dan NBR perlu dilakukan. Berdasarkan analisis GPR dan NBR, model yang layak untuk digunakan adalah model dengan kombinasi variabel prediktor jumlah bayi lahir rendah (X7), jumlah ibu hamil mendapat imunisasi Td2+ (X8), dan jumlah tenaga kesehatan (X10). Keseluruhan variabel berpengaruh signifikan terhadap model. Diperoleh hasil bahwa metode yang paling baik digunakan untuk memodelkan jumlah kematian bayi untuk mengatasi overdispersi adalah metode GPR karena memiliki kriteria kebaikan model AIC, AICc, BIC, dan BICc yang lebih kecil dibandingkan dengan metode NBR.
=================================================================================================================================
Data from the Probolinggo District Health Office in 2020 stated that the Infant Mortality Rate (IMR) in Probolinggo Regency was 8.11. This figure was relatively high compared to the IMR of East Java Province, which was 6.3. Therefore, it is necessary to conduct an analysis to determine the factors that influence the number of infant deaths in Probolinggo Regency in 2020\. Poisson regression is an appropriate method for modeling discrete data. However, Poisson regression requires an equidispersion condition, which is often difficult to meet. In general, discrete data are often characterized by a variance that is greater than the mean, a condition known as overdispersion. Methods that can be used to overcome overdispersion include Generalized Poisson Regression (GPR) and Negative Binomial Regression (NBR). The object of this study consisted of a response variable, namely the number of infant deaths, and predictor variables that were suspected to influence the number of infant deaths, including maternal and infant health factors, hygiene and nutrition factors, and factors related to increasing infant immunity, consisting of 9 variables, as well as 2 additional variables obtained from publications by the Probolinggo Regency BPS, namely the number of health workers and the number of health facilities. The results of the analysis showed that there was an overdispersion condition; therefore, GPR and NBR analyses were conducted. Based on the GPR and NBR analyses, the appropriate model was a model with a combination of predictor variables, namely the number of low-birth-weight infants (X7), the number of pregnant women receiving Td2+ immunization (X8), and the number of health workers (X10). All variables had a significant effect on the model. The results showed that the best method for modeling the number of infant deaths and addressing overdispersion was GPR because it had smaller AIC, AICc, BIC, and BICc values compared with NBR.

Item Type:	Thesis (Other)
Additional Information:	519.536 Cha p-1
Uncontrolled Keywords:	Generalized Poisson Regression, Kematian Bayi, Negative Binomial Regression, Overdispersi, Infant Mortality, Generalized Poisson Regression, Negative Binomial Regression, Overdispersion
Subjects:	Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression
Divisions:	Faculty of Vocational > 49501-Business Statistics
Depositing User:	magang .
Date Deposited:	05 Oct 2026 06:59
Last Modified:	05 Oct 2026 06:59
URI:	http://repository.its.ac.id/id/eprint/145264

Actions (login required)

View Item