Deteksi dan Penanganan Outlier Pada Persamaan Simultan Panel Spasial Autoregressive

Ismadyaliana, Suci (2025) Deteksi dan Penanganan Outlier Pada Persamaan Simultan Panel Spasial Autoregressive. Doctoral thesis, Institut Teknologi Sepuluh Nopember.

Text 7003211009-Dissertation.pdf - Accepted Version Restricted to Repository staff only until 1 April 2027. Download (4MB) | Request a copy

Abstract

Outlier sering kali tersembunyi dalam data, sehingga sulit terdeteksi secara kasat mata. Deteksi outlier penting dalam sebuah pemodelan untuk memastikan bahwa model yang dihasilkan bersifat reliabel. Secara umum, penanganan outlier dapat dilakukan melalui dua pendekatan, yaitu penghapusan outlier (case deletion) dan akomodasi outlier (accommodating). Kerangka Variance Shift Outlier Model (VSOM) mampu mendeteksi sekaligus mengakomodasi outlier dalam model linear dan mixed models. Penelitian sebelumnya juga membuktikan bahwa kerangka Spatial Variance Shift Outlier Model (SVSOM) efektif untuk mengakomodasi outlier pada general spatial models. Keunggulan VSOM terletak pada kemampuannya dalam mengakomodasi baik satu maupun beberapa outlier secara simultan. Deteksi outlier dalam persamaan tunggal telah banyak diteliti, tetapi penelitian mengenai deteksi outlier dalam sistem persamaan simultan masih terbatas. Mengingat minimnya penelitian terkait outlier pada persamaan simultan spasial dengan data panel, penelitian ini mengembangkan metode deteksi dan akomodasi outlier pada persamaan simultan spasial autoregressive data panel menggunakan kerangka SVSOM. Metode ini diterapkan pada sistem persamaan simultan yang melibatkan Foreign Direct Investment (FDI) dan pertumbuhan ekonomi di negara-negara kawasan ASEAN-China Free Trade Area (ACFTA). Hasil penelitian ini menunjukkan bahwa model SVSOM efektif dalam mengatasi outlier pada persamaan simultan spasial autoregressive dengan data panel. Sum of Square Error (SSE) pada model SVSOM jauh lebih kecil dibandingkan model null. Penentuan cut-off untuk deteksi outlier yang dikembangkan lebih mendekati distribusi Chi-square dengan derajat bebas satu dibandingkan dengan metode sebelumnya, dan metode bootstrap parametrik lebih konsisten dalam menghasilkan nilai cut-off dibandingkan dengan metode bootstrap residual.
==================================================================================================================================
Outliers are often hidden within data, making them difficult to detect visually. Detecting outliers is crucial in modeling to ensure the reliability of the resulting model. In general, outliers can be handled using two approaches: outlier deletion and outlier accommodation. The Variance Shift Outlier Model (VSOM) framework was able detect and accommodate outliers in both linear and mixed models. A previous study also demonstrated that the Spatial Variance Shift Outlier Model (SVSOM) framework was effective for accomodating outliers in general spatial models. A key advantage of VSOM is its ability to accommodate both single and multiple outliers simultaneously. While outlier detection in single-equation models has been extensively studied, research on outlier detection in simultaneous equation systems remains limited. Given the scarcity of studies on outliers in spatial simultaneous equations with panel data, this study develops a method for detecting and accommodating outliers in spatial autoregressive simultaneous equations with panel data using the SVSOM framework. This method is applied to a system of simultaneous equation involving Foreign Direct Investment (FDI) and economic growth in ASEAN-China Free Trade Area (ACFTA) countries. The findings confirm that the SVSOM model is effective in addressing outliers in spatial autoregressive simultaneous equations with panel data. The Sum of Square Errors (SSE) in the SVSOM model is significantly lower than in the null model. The cut-off determination for outlier detection developed in this study is closer to the Chi-square distribution with one degree of freedom compared to previous methods. Additionally, the parametric bootstrap method proves to be more consistent in generating cut-off values than the residual bootstrap method.

Item Type:	Thesis (Doctoral)
Uncontrolled Keywords:	accommodating outlier, FDI, persamaan simultan, pertumbuhan ekonomi, SVSOM, economic growth, simultaneous equation
Subjects:	H Social Sciences > H Social Sciences (General) H Social Sciences > HA Statistics H Social Sciences > HA Statistics > HA30.6 Spatial analysis H Social Sciences > HA Statistics > HA31.7 Estimation
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49001-(S3) PhD Thesis
Depositing User:	Suci Ismadyaliana
Date Deposited:	06 Feb 2025 03:25
Last Modified:	06 Feb 2025 03:39
URI:	http://repository.its.ac.id/id/eprint/118421

Actions (login required)

View Item