•  
  •  
 

Abstract

This research addresses the issue of near-multicollinearity within the nonlinear regression framework, specifically the multiple logistic regression model, where the dependent variable is qualitative and binary (representing the occurrence or non-occurrence of a response). The study utilizes Iterative Principal Component Estimators (IPCE) based on both standard weights and conditional Bayesian weights. These estimators were applied to analyze the effects of two drug concentrations, Ciprodar and Garaycin, on patients with pyelonephritis (kidney inflammation), where the dependent variable reflects the clinical outcome (recovery vs. non-recovery). Based on the Mean Squared Error (MSE) criterion, the results demonstrate that the Iterative Principal Component Estimators employing conditional Bayesian weights outperform those relying on standard weights, offering higher precision in the presence of multicollinearity.

DOI

10.33095/jeas.v24i109.1566

Subject Area

Statistical

First Page

535

Last Page

544

Share

COinS