Pemodelan Kasus Pneumonia pada Balita di Kota Bandung Menggunakan Geographically Weighted Panel Regression

Abstract

Abstract. In research, it is not enough to observe only one unit in a period of time, but also to observe in several periods of time. To accommodate this problem, a panel regression method was developed. On the other hand, sometimes the environmental and geographical conditions of one location are different from other locations which then causes spatial heterogeneity. For the case of spatial heterogeneity, a geographically weighted regression analysis was developed, namely Geographically Weighted Regression (GWR). Based on the advantages of the two methods, the Geographically Weighted Panel Regression (GWPR) method was developed which is a combination of the panel regression model with GWR. The purpose of this study was to use the GWPR method to obtain a model and determine the factors that influence pneumonia cases in children under five in each sub-district in Bandung City in 2017-2019. Based on the results of the research, the best panel data regression estimation model is the Fixed Effect Model. The fixed effect GWPR model with adaptive bisquare weighting is the best model because it has the smallest AIC value. The results of testing the suitability of the model with the fixed effect GWPR produce a goodness of fit that is better than the fixed effect panel regression model. Therefore, the GWPR model is useful for modeling cases of pneumonia under five in the district of Bandung City. In the partial test, different models were produced at each location, based on the similarity of the variables that affect each sub-district, 6 groups were obtained.

Keywords: Panel Regression, GWR, GWPR, Pneumonia.

 

Abstrak. Dalam penelitian tidak cukup hanya mengamati satu unit dalam satu periode waktu, tetapi juga perlu mengamati dalam beberapa periode waktu. Untuk mengakomodasi masalah tersebut dikembangkan metode regresi data panel. Di sisi lain terkadang kondisi lingkungan dan geografis satu lokasi berbeda dengan lokasi lainnya yang kemudian menyebabkan heterogenitas spasial. Untuk kasus heterogenitas spasial dikembangkan analisis regresi terbobot geografis yaitu Geographically Weighted Regression (GWR). Berdasarkan kelebihan yang terdapat pada kedua metode tersebut, berkembanglah metode Geographically Weighted Panel Regression (GWPR) merupakan gabungan antara model regresi panel dengan GWR. Tujuan penelitian ini adalah menggunakan metode GWPR untuk mengetahui faktor-faktor yang mempengaruhi kasus pneumonia pada balita tiap kecamatan di Kota Bandung tahun 2017-2019. Berdasarkan hasil penelitian model regresi data panel yang terbaik yaitu Fixed Effect Model. Model fixed effect GWPR dengan pembobot adaptive bisquare merupakan model yang terbaik karena memiliki nilai AIC terkecil. Hasil pengujian kesesuaian model dengan fixed effect GWPR menghasilkan goodness of fit yang lebih baik daripada model fixed effect regresi panel. Karena itu, model GWPR berguna untuk memodelkan kasus pneumonia balita di kecamatan Kota Bandung. Pada uji parsial dihasilkan model yang berbeda pada setiap lokasi, berdasarkan kesamaan variabel yang berpengaruh di tiap kecamatan didapatkan pengelompokkan sebanyak 6 kelompok.

Kata kunci :  Regresi Panel, GWR, GWPR, Pneumonia.