Geographically And Temporally Weighted Log Normal Regression (Studi Kasus: Indeks Keparahan Kemiskinan Provinsi Jawa Timur)

Hidayanty, Nurul Ilma (2022) Geographically And Temporally Weighted Log Normal Regression (Studi Kasus: Indeks Keparahan Kemiskinan Provinsi Jawa Timur). Masters thesis, Institut Teknologi Sepuluh Nopember Surabaya.

Text 6003202005-Master_Thesis.pdf - Accepted Version Download (3MB)

Abstract

Log Normal Regression (LNR) merupakan model regresi dengan variabel respon yang mengikuti distribusi Log Normal. Pemodelan ini menghasilkan penaksiran parameter yang sifatnya global untuk seluruh wilayah pengamatan. Penggunaan model LNR pada data panel dengan unit pengamatan berupa wilayah kurang tepat digunakan karena memungkinkan adanya heterogenitas secara spasial-temporal. Penelitian ini berfokus untuk mengembangkan model Geographically and Temporally Weighted Log Normal Regression (GTWLNR) untuk menganalisa data dengan dua karakteristik utama yakni variabel respon berdistribusi Log Normal dan mempertimbangkan heterogenitas spasial-temporal. Penaksiran parameter model GTWLNR menggunakan Maximum Likelihood Estimation (MLE), apabila hasil yang diperoleh tidak closed form maka diselesaikan dengan iterasi numerik Newton Raphson. Pengujian hipotesis untuk uji simultan menggunakan metode Maximum Likelihood Ratio Test (MLRT). Pengujian parsial di 38 kabupaten/kota selama lima periode menunjukkan hasil signifikansi variabel prediktor yang berbeda-beda pada setiap tahunnya. Variabel signifikan dikelompokkan menjadi 16 kelompok, dimana daerah yang berdekatan cenderung memiliki signifikansi yang sama. Variabel persentase rumah tangga yang memiliki sumber air minum yang layak dan persentase rumah tangga yang memiliki sanitasi yang layak signifikan hampir disemua kabupaten/kota di Jawa Timur pada tiap periode, sedangkan keempat variabel lainnya memiliki signifikansi yang bervariasi antar lokasi dan periode. Ukuran kebaikan model menggunakan Akaike's Information Criterion (AIC). Berdasarkan hasil AIC dan AICc, model lokal lebih baik digunakan karena menghasilkan nilai AIC dan AICc lebih kecil jika dibandingkan model global
====================================================================================================================================
Log Normal Regression (LNR) is a regression model with a response variable that follows the Log Normal distribution. This modeling results in parameter estimates that are global for the entire observation area. The use of the LNR model on panel data with the observation unit an area is not appropriate because it allows for spatial-temporal heterogeneity. This study focuses on developing a Geographically and Temporally Weighted Log Normal Regression (GTWLNR) model to analyze data with two main characteristics, the response variable with a Log Normal distribution and considering spatial-temporal heterogeneity. The parameter estimation of the GTWLNR model uses the Maximum Likelihood Estimation (MLE), if the results obtained are not closed form then it is solved by Newton Raphson numerical iteration. Hypothesis testing for the simultaneous test uses the Maximum Likelihood Ratio Test (MLRT) method. Partial testing in 38 regencies/cities for five periods showed different significance of predictor variables each year. Significant variables were grouped into 16 groups, where adjacent areas tended to have the same significance. The variable percentage of households that have adequate drinking water sources and the percentage of households that have proper sanitation is significant in almost all districts/cities in East Java in each period, while the other four variables have significance that varies between locations and periods. The measure of the goodness of the model uses Akaike's Information Criterion (AIC). Based on the results of AIC and AICc, the local model is better to use because it produces smaller AIC and AICc values when compared to the global model

Item Type:	Thesis (Masters)
Additional Information:	RTSt 519.536 Hid g-1 2022
Uncontrolled Keywords:	GTWLNR; Indeks Keparahan Kemiskinan; MLE; MLRT; GTWLNR; Poverty Severity Index; MLE; MLRT
Subjects:	Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49101-(S2) Master Thesis
Depositing User:	EKO BUDI RAHARJO
Date Deposited:	28 Mar 2023 02:41
Last Modified:	18 Nov 2024 01:12
URI:	http://repository.its.ac.id/id/eprint/97815

Actions (login required)

View Item