Strange results for AUC and Gini

[Vasisualiy Lokhankin]

Hi,

I estimated a logistic regression, generated predicted values, and then calculated AUC and Gini index as measures of the model's performance. I obtained AUC = 0.6638 and Gini index = 0.6034. Oddly enough, it does not hold that Gini = 2AUC-1. Why it is so?

I simply used commands

roctab refvar pd, summary
roctab refvar pd, lorenz

The first one gives AUC, the second one gives Gini.

I attach a dataset with values of the reference variable and predicted values.

Thanks in advance.