Assessment of Influential Points Detection in Gamma-Pareto Regression Residuals Using Diagnostic Measure, Difference of Fits

The assessment of Influential points detection in gamma-pareto regression

Abstract

The principal diagnosis process in order to achieve reliable output of the regression is the influential analysis. Same applies in case of the generalized linear model. In the present paper, the practical comparison of the functioning of different residuals of the Gamma-Pareto regression model is made in order to define the points of influence. The residuals of Gamma-Pareto regression model are further divided into two i.e., standardized and adjusted residuals. The said two residuals have both undergone difference of fits and consequently comparison of these residuals in the finding of influential points were carried out by simulation and Ardennes related data set. Simulation result shows that in case dispersion parameter is very low, likelihood residuals will perform better compared to the others and all of the adjustment type of residuals will perform identically but not superior to the standardized. Although, in case of large values of dispersion parameter, all the standardized residuals behave similarly, and they are superior to likelihood residuals in revealing the presence of influential points.

Dhaka Univ. J. Sci. 74(1): 138-159, 2026 (January)