Table 2.

Results of simulation study showing the selection bias in estimating an association that is not null in the intended study population (regression coefficient for outcome on phenotype is 0.1)

Simulation settings		Results—association between allele score and outcome
Association between missingness and both phenotype and outcome (OR)	Association between allele score and phenotype (r)	Mean regression coefficient (SD)	True regression coefficient	Number of 95% CIs containing true value
OR = 1.8	0.05	0.003 (0.001)	0.005	78
	(0.25% variance)
	0.10	0.006 (0.001)	0.01	23
	(1.00% variance)
	0.15	0.010 (0.001)	0.015	2
	(2.25% variance)
	0.20	0.013 (0.001)	0.02	0
	(4.00% variance)
	0.30	0.020 (0.001)	0.03	0
	(9.00% variance)
OR = 1.5	0.05	0.004 (0.001)	0.005	94
	(0.25% variance)
	0.10	0.009 (0.001)	0.01	86
	(1.00% variance)
	0.15	0.013 (0.001)	0.015	69
	(2.25% variance)
	0.20	0.017 (0.001)	0.02	53
	(4.00% variance)
	0.30	0.026 (0.001)	0.03	19
	(9.00% variance)
OR = 1.2	0.05	0.005 (0.001)	0.005	98
	(0.25% variance)
	0.10	0.01 (0.001)	0.01	96
	(1.00% variance)
	0.15	0.014 (0.001)	0.015	94
	(2.25% variance)
	0.20	0.019 (0.001)	0.02	92
	(4.00% variance)
	0.30	0.029 (0.001)	0.03	95
	(9.00% variance)

Each scenario was simulated 100 times.