Estimating Parameters of the Beta Regression Model Using the Dung Beetle Optimizer: Comparative Analysis with BFGS and Application to Thalassemia Data
Adnan Mostafa Al-Sinjary *
Department of Statistics and Informatics Techniques, College of Administrative Technology, Northern Technical University, Iraq.
*Author to whom correspondence should be addressed.
Abstract
The beta regression model is used for data that is continuous but confined between zero and one, such as percentages. To study these models, it is necessary to estimate their parameters using common estimation methods, primarily the Maximum Likelihood Estimator (MLE). In this research, we relied on a modern metaheuristic algorithm, the Dung Beetle Optimizer (DBO), and compared its behavior with a conventional method, the BFGS algorithm. Simulations were conducted to compare the two methods using the Mean Squared Error (MSE) criterion, the results showed that DBO provided more accurate estimates, particularly in small-sample and high-precision scenarios. Furthermore, the impacts of factors including age, hemoglobin level, and the quantity of blood units on the packed cell volume (PCV) were investigated using the Beta regression model on actual data from thalassemia patients. The application demonstrated how well the model handled proportional data with a 0–1 boundary and demonstrated how important some variables were in describing the event under study.
Keywords: Beta regression model, metaheuristic algorithms, BFGS algorithm, dung beetle optimizer