Table 3.

Proportion of cases where each method fails to select any features is shown along with averages of operating characteristics calculated across 500 simulations for each of 6 methods and 5 simulation scenarios, SI-SV (described in the text). Standard errors are order 0.001. The method with the highest average validation performance is highlighted.

Method
Screening:	–	SIS	Lasso	Average	Index	Index
Final Model:	Lasso	Lasso	Lasso	Lasso	Count	Lasso
Proportion of cases when method fails to select any features
SI	0.17	0.02	0.17	0.48	0.00	0.00
SII	0.05	0.00	0.19	0.00	0.00	0.00
SIII	0.11	0.00	0.14	0.00	0.00	0.00
SIV	0.05	0.00	0.12	0.45	0.00	0.00
SV	0.17	0.01	0.18	0.59	0.00	0.00
Correlation between true and predicted ranks
SI	0.26	0.21	0.18	0.00	0.13	0.20
SII	0.26	0.33	0.43	0.99	0.32	0.46
SIII	0.16	0.23	0.30	0.84	0.29	0.33
SIV	0.39	0.36	0.41	0.02	0.37	0.45
SV	0.17	0.20	0.23	0.01	0.23	0.24
Manhattan distance relative to all same rank.
SI	0.56	0.58	0.60	0.67	0.61	0.58
SII	0.56	0.53	0.48	0.06	0.53	0.47
SIII	0.60	0.57	0.54	0.25	0.54	0.53
SIV	0.50	0.51	0.49	0.66	0.51	0.47
SV	0.59	0.58	0.57	0.66	0.56	0.56
Misclassification in 10% tails
SI	0.29	0.32	0.38	0.50	0.40	0.36
SII	0.29	0.23	0.20	0.00	0.27	0.18
SIII	0.37	0.31	0.29	0.01	0.29	0.27
SIV	0.19	0.21	0.21	0.48	0.23	0.18
SV	0.35	0.33	0.34	0.50	0.34	0.33