A statistical preprint combines clustering, regression and missing-data treatment to identify unusual cases in a 205-car dataset.