TABLE 3.

Comparison of proposed methods with misspecified models (Simulation Study 2): Time between study visits ($T_V$)

Scenario	%No disease	% Treated disease	% Untreated disease	Method	$T_M^a$	$Y_V^a$	Bias	Relative Bias	Empirical SE	Estimated SE	Coverage
1	50	15	35	IPW Proposed	T		0.020	0.010	0.390	0.375	94.1
					F		−0.091	−0.043	0.356	0.348	93.5
				DR Proposed	T	T	0.015	0.007	0.318	0.323	94.6
					T	F	0.020	0.009	0.322	0.325	95.0
					F	T	0.013	0.006	0.318	0.323	94.8
					F	F	−0.081	−0.038	0.314	0.318	94.5
2		25	25	IPW Proposed	T		−0.010	−0.005	0.522	0.489	93.0
					F		−0.203	−0.098	0.370	0.378	91.2
				DR Proposed	T	T	0.024	0.011	0.324	0.324	93.4
					T	F	0.038	0.018	0.345	0.338	93.5
					F	T	0.019	0.009	0.322	0.322	94.2
					F	F	−0.137	−0.066	0.315	0.315	91.9
3	70	15	15	IPW Proposed	T		−0.025	−0.020	0.522	0.486	92.4
					F		−0.128	−0.102	0.370	0.370	93.1
				DR Proposed	T	T	0.002	0.001	0.298	0.298	95.0
					T	F	0.003	0.003	0.306	0.306	94.7
					F	T	0.010	0.008	0.296	0.297	95.0
					F	F	−0.080	−0.064	0.288	0.290	93.9
4		25	5	IPW Proposed	T		−1.534	−1.180	0.984	0.900	54.9
					F		−0.512	−0.394	0.717	0.664	81.6
				DR Proposed	T	T	−0.001	−0.000	0.306	0.304	94.2
					T	F	−0.236	−0.182	0.320	0.322	88.2
					F	T	−0.004	−0.003	0.299	0.301	94.1
					F	F	−0.175	−0.134	0.285	0.288	91.4

Abbreviations: SE, standard error.

Note: The true population parameter was calculated based on fitting the model of interest in the population dataset.

^a$T$ = The model was specified correctly. $F$ = The model was misspecified by excluding the auxiliary variable $Z$.