Comparison of Some Methods for Estimating Nonparametric Binary Logistic Regression

Abstract

In this research, the methods of Kernel estimator (nonparametric density estimator) were relied upon in estimating the two-response logistic regression, where the comparison was used between the method of Nadaraya-Watson and the method of Local Scoring algorithm, and optimal Smoothing parameter λ was estimated by the methods of Cross-validation and generalized Cross-validation, bandwidth optimal λ has a clear effect in the estimation process. It also has a key role in smoothing the curve as it approaches the real curve, and the goal of using the Kernel estimator is to modify the observations so that we can obtain estimators with characteristics close to the properties of real parameters, and based on medical data for patients with chronic lymphocytic leukemia and through the use of the Gaussian function and based on the comparison criterion (MSE) it was found that the Nadaraya -Watson method is the best because it obtained the lowest value for this criterion.