A Comparison of a General Linear Model and the Ratio Estimator

Abstract

In data analysis, selecting a proper statistical model is a challenging issue. Upon the selection, there are other important factors impacting the results. In this article, two statistical models, a General Linear Model (GLM) and the Ratio Estimator will be compared. Where applicable, some issues such as heteroscedasticity, outliers, etc. and the role they play in data analysis will be studied. For reducing the severity of heteroscedasticity, Weighted Least Square (WLS), Generalized Least Square (GLS), and Feasible Generalized Least Square (FGLS) will be deployed. Also, a revised version of FGLS is introduced. Since these issues are data dependent, shrimp effort data collected in the Gulf of Mexico for the years 2005 through 2018 will be used and it is shown that the revised FGLS reduces the impact of heteroscedasticity significantly compared to that of FGLS. The data sets will also be checked for the outliers and corrections are made (where applicable). It is concluded that these issues play a significant role in data analysis and must be taken seriously. Further, the two statistical models, that is, the GLM and the Ratio Estimator are compared.

Analisis Performansi Algoritma Linear Regression dengan Generalized Linear Model untuk Prediksi Penjualan pada Usaha Mikra, Kecil, dan Menengah

“…Model ini juga merupakan perpanjangan dari variabel dependen diperbolehkan memiliki distribusi yang berbeda [25]. Perspektif yang berbeda menyebabkan beberapa generalisasi pada koefisien penentuan [26].…”