PEMERIKSAAN DATA BERPENGARUH DALAM MODEL REGRESI GAMMA

Abstract

In certain cases, often encountered a group of data where the observed respone variable shaped non-negative and not symmetrical (skewed to the right). The case data such as this one can be found in the field of insurance, such as variable stating the amount of claims or states of time between when a customer claims to obtain these claims. To handle such cases one of them is by using a generalized linear model, with respone variables that follow the gamma distribution. The paper discusses the inspection data outliers and influential data in the modeling of the respone following the gamma distribution. Several statistical measures used to examine the outlier data is the value of leverage, standardized deviance residual, standardized Pearson residual, and residual likelihood. Then the data outliers potentially influential data will be examined using Cook's distance statistics.

Keywords: Gamma Distributions, Leverage, Standardized Deviance Residual, Standardized Pearson Residual, And Residual Likelihood, Cook's Distance Statistics