Adjusted \(R^2\) measures for the inverse Gaussian regression model (Q1424619)

The paper presents two different definitions of the \(R^2\) measure (coefficient of determination) for the inverse Gaussian regression model. The first definition is based on deviances and the second one is based on the sums-of-squares principle. Proper adjustments for the proposed \(R^2\) measures necessary to reduce their bias are proposed and the corresponding adjusted \(R^2\) measures are investigated and compared in a simulation study. Their application is illustrated by a real medicine data example. The use of adjusted \(R^2\) measures is recommended in general.