Table 2.

Nested case–control study simulation results showing convergence with increasing sample size, iterations = 1000, perturbations = 500, = 0.1, and = 1.0

Nested case–control, iterations = 1000, perturbations = 500, = 0.1, = 1.0
	Sampled from n = 500, s = 48, = 24, = 58/27/11
	True	Est	Bias %	SE	SE	CP %
PE	0.168	0.171	1.6	0.021	0.021	95.5
TPF(0.4)	0.936	0.929	0.7	0.044	0.046	91.2
FPF(0.3)	0.777	0.767	1.2	0.107	0.105	91.2
AUC	0.813	0.780	4.1	0.050	0.055	95.2
PCF(0.2)	0.297	0.306	2.8	0.026	0.032	98.0
PNF(0.8)	0.640	0.650	1.6	0.051	0.055	94.1
	Sampled from n = 4000, s = 48, = 24, = 462/219/85
PE	0.168	0.169	0.6	0.007	0.007	96.4
TPF(0.4)	0.936	0.935	0.1	0.015	0.015	94.1
FPF(0.3)	0.777	0.777	0.0	0.037	0.037	94.2
AUC	0.813	0.806	0.8	0.017	0.017	95.7
PCF(0.2)	0.297	0.298	0.2	0.009	0.009	94.8
PNF(0.8)	0.640	0.642	0.3	0.017	0.018	94.6

PE = prediction error; TPF(0.4) = true positive fraction at ; FPF(0.3) = false positive fraction at ; AUC = area under the ROC curve; PCF(0.2) = the proportion of events in timeframe from captured if 20% of the population at risk at time were followed; PNF(0.8) = the proportion of the population at risk at time that would need to be followed in order to capture 80% of the events in the timeframe from . Effective sample sizes used to estimate the performance measures are summarized in the table as , where denotes the number of events observed in a timeframe from , denotes the number of subjects still at risk at time and is the number of subjects who were censored between and .