Regression Diagnostics in Gamma Regression and Partial Residual Plots

Abstract

This study describes partial residual plots with its structure and utility in the generalized linear model (GLM) setting. These plots are used as a tool for visualising diagnostics and curvature as a function of chosen predictors. In this case, partial residual plots are constructed and gamma regression as a GLM is taken into consideration. Depending on how the response variable behaves and how the affiliated link function interacts with various covariates, these graphs may or may not be effective at providing a clear visual representation of curvature and diagnostics. We investigate the behaviour of the population version, the estimated coefficients, and the partial residuals. For improved perception of fit issues, specification issues, and data abnormalities, many diagnostics in a single display are available.