Comparison of Binary Logistic Regression Models Based on Bootstrap Method: An Application on Coronary Artery Disease Data

Abstract

This study is aimed to obtain an appropriate logistic regression model based on the bootstrapmethods. For this purpose, two bootstrap methods called bootstrap I and bootstrap II are given toobtain the estimations of parameters and standard errors. Traditional logistic regression iscompared with the bootstrap I and bootstrap II methods in terms of the parameter estimations andstandard errors. It has been found that the standard errors of the parameter estimations for thebootstrap I model are smaller than others. Also, the average widths of confidence interval basedon bootstrap I model are narrower than the logistic regression and bootstrap II. It is seen that, thesimulation study based on different sample sizes supports these results. It can be said that thebootstrap I model based on resampling of errors term is the best in estimating coronary arterydisease.