. Author manuscript; available in PMC: 2014 Jan 5.

Published in final edited form as: Stat Med. 2011 Aug 4;30(24):10.1002/sim.4322. doi: 10.1002/sim.4322

Table 4.

“Finding A” performance: variations of base-case (θ = 0.90, true A = {X₁ > 0, X₂ < 0}) and null-case (θ = 0 true A null) for VT(R)_0.05.

	Sensitivity	Specificity	AUC	PPV	NPV	Proportion Â null	Median \|Â\|	{X₁, X₂}∈ tree	# Unique X’s
θ = 0.9:
Base Case	0.47	0.89	0.77	0.55	0.84	0.07	189	0.51	3.62
p = 30	0.48	0.89	0.78	0.51	0.85	0.15	213	0.52	2.89
Correlated	0.52	0.85	0.76	0.52	0.85	0.03	247	0.32	4.08
Sub-spec (a_i)	0.54	0.89	0.79	0.60	0.86	0.04	218	0.62	3.59
Sub-spec (a_i, b_i)	0.50	0.89	0.76	0.57	0.85	0.05	216	0.48	3.61
n = 400	0.40	0.82	0.66	0.42	0.81	0.05	93	0.19	4.09
n = 2000	0.68	0.93	0.87	0.74	0.90	0.03	462	0.78	3.03
Larger \|A\|	0.41	0.91	0.78	0.77	0.61	0.06	233	0.54	3.35
θ = 1.5	0.78	0.91	0.89	0.77	0.93	0.00	245	0.83	3.39
θ = 0:
Base Case	-	0.87	-	-	-	0.25	131	-	3.29
p = 30	-	0.88	-	-	-	0.27	103	-	2.90
Correlated	-	0.82	-	-	-	0.09	164	-	4.19
Sub-spec (a_i)	-	0.84	-	-	-	0.17	152	-	3.44
Sub-spec (a_i, b_i)	-	0.86	-	-	-	0.28	131	-	3.25
n = 400	-	0.78	-	-	-	0.05	86	-	4.10
n = 2000	-	0.88	-	-	-	0.25	230	-	2.84