GENERALIZED LINEAR MODEL WITH ROBUST PARAMETER ESTIMATOR FOR BINOMIAL CASE

• Main Author: Rafly Keliat, Erick
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/49719
• Institution: Institut Teknologi Bandung

Generalized Linear Model is one of many kinds of modelling in Statistics. Commonly, linear regression is often used and recognized widely. However, Generalized Linear Model is the generalized form of the linear regression itself and it makes many data types can be used in the modelling. One of many problems that faced in data modelling is the existence of outliers. Outliers give quite bad effects in the model formulation, but they also give several important information. Therefore, robust models are needed to overcome this problem. In this Thesis, a simulation and a case study of two different data are conducted with maximum likelihood estimator, MT estimator, and WMT estimator. The aim is to determine the robustness of each estimator. The simulation result shows that MT estimator and WMT estimator are more robust than maximum likelihood estimator based on their MSE values. The case study result shows that MT estimator and maximum likelihood estimator give the best robust models based on its boxplots of the absolute values of the deviance residuals.