The case-crossover design via penalized regression

Table 2 Area under the curve for binary discrimination of drugs

	a) AUC for “highly suspected drugs“	b) AUC for “highly suspected drugs“ or “suspected drugs“
Univariable CLR	0.836	0.871
Multivariable CLR	0.741	0.775
Lasso	0.862	0.861
Elastic net	0.868	0.909
Bolasso	0.804	0.813
Sublasso	0.815	0.823
Random lasso	0.820	0.847
Firth correction	0.852	0.928
Case-control^a	0.947	0.809

ISSN: 1471-2288