. Author manuscript; available in PMC: 2015 Sep 28.

Published in final edited form as: Stat Med. 2014 May 13;33(22):3946–3959. doi: 10.1002/sim.6195

Table 3.

Estimates for the preliminary ROC curve for Model 1. Using r_j and w_j, we compute fpr_j and tpr_j and hence ROCSlope_j. The preliminary ROC curve is not concave because ROCSlope_j decreases from j=8 to j=9 (from right to left on the ROC curve).

Model 1 risk interval		x_j	n_j	w_j	r_j	fpr_j	tpr_j	ROCSlope_j
0.00–0.10	j = 1	95	1500	0.2647	0.0633	1	1	0.1734
0.10– 0.20	j = 2	52	356	0.0628	0.1461	0.6554	0.9403	0.4386
0.20– 0.30	j = 3	225	906	0.1599	0.2483	0.5808	0.9075	0.8472
0.30– 0.40	j = 4	497	1424	0.2513	0.3490	0.4138	0.7660	1.3747
0.40– 0.50	j = 5	452	1019	0.1798	0.4436	0.1864	0.4535	2.0441
0.50– 0.60	j = 6	196	347	0.0612	0.5648	0.0473	0.1692	3.3283
0.60– 0.70	j = 7	57	96	0.0169	0.5938	0.0103	0.0459	3.7476
0.70– 0.80	j = 8	15	17	0.0030	0.8824	0.0007	0.0101	19.2311
0.80– 0.90	j = 9	1	2	0.0004	0.5000	0.0002	0.0006	2.5642
						0	0