Table 4.

Percent median bias for estimation of β_S and β_SR in model (4), under various design/estimator combinations. Estimates are based on B = 10,000 simulated datasets.

Design	Sex main effect, βS			Interaction, βSR
	WL	PL	ML	WL	PL	ML
Complete data	-	-	0.0	-	-	0.6
Case-control designa	-	-	1.7	-	-	38.7
Two-phase design
Sexb	1.4	0.1	0.1	5.7	2.7	1.1
Raceb	6.3	2.9	2.9	−15.5	1.4	1.4
Sex and Race jointlyc	0.9	−0.4	−0.4	−5.7	0.6	0.6

^abalanced case-control design with n₀ = n₁ = 100,

^bBalanced two-phase design with n_yk = 50 for y = 0,1 and k = 0,1,

^cBalanced two-phase design with n_yk = 25 for y = 0,1 and k = 0,1, 2, 3