The models are fit using maximum rank correlation (MRC) method which
is an alternate approach to logistic regression. In MRC the area under
the ROC curve (AUC) is maximized as opposed to the likelihood in
logistic regression. Due to invariance of AUC to location and scale
shifts one of the parameters (anchor variable) is set to 1.
The first variable (column) in x is used as the anchor variable.
The IPMN data set used as an example in the paper below is included.
The columns are high risk lesion (V1), recent weight loss (V2), main
duct involvement (V4), presence of a solid component in imaging (V3),
and lesion size (V5).