Table 5.

Operating characteristics, by the phase II sample size, of four proposed estimators in the setting where Z is discrete. Results are based on a series of simulations each consisting of 20,000 repetitions.

Phase II sample size	Scheme/Estimator	Percent bias				Relative efficiency
		Intercept	Gender	Weight	Gestation	Intercept	Gender	Weight	Gestation
400	Full participation
	WL	2.4	7.7	6.3	36.0	257	332	216	218
	PL	0.1	0.1	2.1	3.9	100	100	100	100
	Selective participation
	WL
	Näive	5.1	13.4	11.8	51.3	295	375	242	248
	FWL—estimated	2.4	8.8	6.2	36.5	258	335	220	221
	PL
	Näive	1.6	0.8	3.6	35.3	118	122	109	116
	WPL—estimated	0.2	−0.4	2.6	3.7	111	126	111	106
	EPL	0.2	0.7	2.5	3.3	101	96	109	104
	MO
	M = 5	0.2	0.6	2.9	3.0	103	96	112	106
	M = 10	0.2	0.6	2.9	3.0	102	95	111	106
1000	Full participation
	WL	1.0	4.4	1.5	15.1	250	319	216	228
	PL	0.1	0.5	1.0	1.3	100	100	100	100
	Selective participation
	WL
	Näive	3.0	4.8	6.5	24.8	282	354	238	253
	FWL—estimated	1.0	1.8	1.3	15.0	251	321	216	229
	PL
	Näive	1.6	1.3	2.1	31.8	116	120	107	114
	WPL—estimated	0.1	0.5	1.1	1.1	110	123	108	105
	EPL	0.1	0.4	1.1	0.9	100	95	107	103
	MO
	M = 5	0.1	0.4	1.2	0.8	101	95	109	105
	M = 10	0.1	0.4	1.2	0.8	101	95	108	105