Title (Arabic)
تقدير النموذج اللوجستي باستخدام اوزان بيز المتسلسل
DOI
10.33095/jeas.v13i46.1279
Abstract
This paper addresses the critical statistical challenge of heteroscedasticity—the non-constant variance of random error—within the framework of Multiple Logistic Regression. When the response variable is qualitative or binary, traditional estimation methods often struggle to maintain efficiency under varying error variances. To resolve this, the authors propose an innovative two-step sequential estimation procedure that integrates the Sequential Bayesian (SB) approach with Weighted Least Squares (WLS) techniques. Unlike classical methods that treat observations in a fixed batch, the sequential Bayesian mechanism allows for the dynamic updating of parameter estimates as observations are processed one by one, effectively incorporating new information into the model's posterior distribution. The research utilizes an Empirical Bayes concept to derive the weights necessary for the WLS estimation, thereby stabilizing the model against heteroscedasticity. Through a practical application involving clinical data—specifically the recovery of patients from severe acute renal conditions treated with different antibiotics—and a MATLAB-based simulation, the study demonstrates that the Sequential Bayesian Weighted approach significantly outperforms classical methods. The performance is validated using statistical criteria such as Mean Square Error (MSE) and the coefficient of determination (R^2), proving that the proposed model provides more accurate and reliable parameter estimates, especially in complex or small-sample scenarios.
Abstract (Arabic)
تقدير النموذج اللوجستي باستخدام اوزان بيز المتسلسل
Recommended Citation
Ali, T. H., & Hassan Chawsheen, T. A. (2007). Logistic Model Estimation by Sequential Bayesian Weights. Journal of Economics and Administrative Sciences, 13(46). https://doi.org/10.33095/jeas.v13i46.1279
