Robust Logistic Modelling for Datasets with Unusual Points

Authors : Kumru URGANCI TEKIN, Burcu MESTAV, Neslihan İYİT

Pages : 49-63

Doi:10.53570/jnt.971062

View : 13 | Download : 8

Publication Date : 2021-09-30

Article Type : Research Paper

Abstract :Unusual Points insert ignore into journalissuearticles values(UPs); occur for different reasons, such as an observational error or the presence of a phenomenon with unknown cause. Influential Points insert ignore into journalissuearticles values(IPs);, one of the UPs, have a negative effect on parameter estimation in the Logistic Regression model. Many researchers in fisheries sciences face this problem and have recourse to some manipulations to overcome this problem. The limitations of these manipulations have prompted researchers to use more suitable and innovative estimation techniques to deal with the problem. In this study, we examine the classification accuracies and parameter estimation performances of the Maximum Likelihood insert ignore into journalissuearticles values(ML); estimator and robust estimators through modified real datasets and simulation experiments. Besides, we discuss the potential applicability of the assessed robust estimators to the estimation models when the IPs are kept in the dataset. The obtained results show that the Weighted Maximum Likelihood insert ignore into journalissuearticles values(WML); and Weighted Bianco-Yohai insert ignore into journalissuearticles values(WBY); estimators of robust estimators outperform the others.
Keywords : Influential point, robust estimators, unusual point, logistic regression