Influence measures in gamma modified ridge type estimator

Multicollinearity and unusual observations pose threats to the performance of maximum likelihood estimator in the gamma regression model. Literature has shown that both problems can exist simultaneously in a model. Researchers have paid little attention to the detection of influential observation in the gamma regression model with multicollinearity. This study aims to develop statistics for the detection of unusual observation in a multicollinear gamma regression model using gamma modified ridge-type estimator. The performance of the statistics was examined through a Monte Carlo simulation study and two real applications. The results show that gamma modified ridge-type estimator copes with unusual observations by reducing their influence.