Table 3:

Results from Scenario 3 (n=20,000 and and ~87% missingness) based on the naïve complete case analysis (CC), the multiple imputation by chained equations approach (MICE), and the proposed sieve inverse probability weighting approach (SIPW).

Analysis		Event 1		Event 2
		${\widehat{\beta }}_{11}$	${\widehat{\beta }}_{21}$	${\widehat{\beta }}_{12}$	${\widehat{\beta }}_{22}$
CC	% Bias	−22.842	−21.349	36.646	37.663
	MCSDa	0.030	0.064	0.047	0.092
	ASEb	0.029	0.064	0.046	0.092
	CPc	0.028	0.091	0.019	0.473
MICE	% Bias	−9.713	−11.929	−30.823	−31.537
	MCSDa	0.032	0.063	0.027	0.050
	ASEb	0.042	0.077	0.028	0.050
	CPc	0.873	0.780	0.011	0.276
SIPW	% Bias	−0.376	−0.546	0.467	0.269
	MCSDa	0.043	0.092	0.052	0.098
	ASEb	0.042	0.090	0.053	0.097
	CPc	0.947	0.938	0.949	0.942

^aMonte Carlo standard deviation of the estimates.

^bAverage of the standard error estimates.

^cEmpirical coverage probability.