Please use this identifier to cite or link to this item: http://repository.ipb.ac.id/handle/123456789/105386
Title: Kajian Perbandingan Metode Least Median of Squares dan Maximum Likelihood Type pada Regresi Kekar
Authors: Sadik, Kusman
Suhaeni, Cici
Mutiarasari, Vina Fauzia
Issue Date: 2021
Publisher: IPB University
Abstract: Keberadaan pencilan dapat memengaruhi pendugaan parameter dalam analisis regresi yang menggunakan metode ordinary least squares (OLS). Metode regresi kekar dapat digunakan sebagai pendekatan alternatif dari OLS. Penelitian ini dilakukan untuk mengkaji perbandingan dari dua metode regresi kekar, yaitu least median of squares (LMS) dan maximum likelihood type (M) melalui data simulasi. Data simulasi diperoleh dengan membangkitkan bilangan acak berdasarkan model regresi linear berganda dua peubah penjelas. Kombinasi yang digunakan adalah jenis pencilan, proporsi pencilan, dan ukuran contoh. Berdasarkan nilai bias mutlak dan kuadrat tengah galat (KTG) diperoleh bahwa metode M lebih tepat diterapkan ketika data mengandung vertical outlier atau good leverage point, lalu diikuti oleh metode LMS. Sementara itu, metode LMS dapat menduga parameter regresi pada data yang mengandung bad leverage point. Analisis selanjutnya dilakukan pada data pertanian mengenai pengaruh pupuk organik dan konsumsi beras per kapita terhadap produksi padi sawah di 34 provinsi tahun 2017. Hasil yang diperoleh menunjukkan bahwa metode LMS menghasilkan nilai bias mutlak dan KTG terkecil.
The presence of outliers can be affect to estimate the parameters of the linear regression using ordinary least squares (OLS). The robust regression methods can be used as an alternative to OLS. This study aims to compare the performance of two methods, namely least median of squares (LMS) and maximum likelihood type (M) using simulation study. The simulation data obtained by generating random number based on multiple linear regression with two explanatory variables. With combination of the sample size, outliers type, and the number of outliers. Based on the criteria: absoulut bias and mean square error (MSE), the M-estimator gave better result than the LMS-estimator when the data contains vertical outliers or good leverage point. Meanwhile, the LMS-estimator gave better result at data containing bad leverage points. The next step is analysed data application for agriculture on the affect of organic fertilizer and rice consumption per capita against the use of rice production in 34 provinces in 2017 using the best method based on data characteristic. The results from LMS-estimator produce the smallest absolut bias and MSE.
URI: http://repository.ipb.ac.id/handle/123456789/105386
Appears in Collections:UT - Statistics and Data Sciences

Files in This Item:
File Description SizeFormat 
Cover, Lembar Pengesahan, Prakata, Daftar Isi.pdf
  Restricted Access
Cover998.53 kBAdobe PDFView/Open
G14160054_VINA FAUZIA MUTIARASARI.pdf
  Restricted Access
Fullteks1.43 MBAdobe PDFView/Open
Lampiran.pdf
  Restricted Access
Lampiran695.66 kBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.