•  
  •  
 

Journal of Economics and Administrative Sciences

Abstract

Influence diagnostics are essential for identifying influential observations that may affect the estimation accuracy of regression models. Classical diagnostic measures such as Cook's Distance and DFFITS may become less effective in complex or highly dispersed datasets. In this study, a new Hybrid Secretary–Osprey Optimization Algorithm (HSOOA), based on the integration of the Secretary Bird Optimization Algorithm (SBOA) and the Osprey Optimization Algorithm (OOA), is proposed for influential observation detection in the Gamma Regression Model (GRM). This optimization framework successfully achieves an optimal balance between global exploration and local exploitation. The proposed HSOOA framework was applied to a real medical dataset related to Fasting Blood Glucose (FBG), where 18 influential observations were successfully identified. Instead of removing these observations, their influence was reduced using a robust M-estimation procedure based on the proposed HSOOA diagnostic weighting function, with the aim of improving the stability and reliability of the Gamma Regression Model estimates. The main advantage of this approach is its capability to bound the severe impact of outliers while preserving all critical clinical data. The value of Mean Squared Error (MSE) obtained after the treatment process was relatively low (specifically MSE = 1.154), which shows the effectiveness of the proposed robust framework in reducing the effect of influential observations and improving the estimation accuracy of the fitted Gamma Regression Model in applied statistical modeling.

DOI

10.33095/2227-703X.4366

Article Type

Research Article

Subject Area

Statistical

First Page

111

Last Page

121

Creative Commons License

Creative Commons Attribution-NonCommercial 4.0 International License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License

Download

Share

COinS