Table 3.

Bias and Monte Carlo standard errors for each estimator in the simulation study. Bias here is defined as the estimate - true value. 1000 repetitions of 1000 independent draws were taken. All estimators are for the ITT treatment strategy parameter except for “standard IPW.”

	$E(Y^{k=2})-E(Y^{k=3})$	$E(Y^{k=3})-E(Y^{k=4})$	$E(Y^{k=4})-E(Y^{k=5})$
Scenario 1: No direct effect of treatment, no effect of delivery time
True value	0	0	0
IPW	$-$ 0.001 (0.042)	$-$ 0.001 (0.041)	0.000 (0.031)
G-computation	0.001 (0.040)	$-$ 0.002 (0.040)	0.001 (0.029)
TMLE	$-$ 0.001 (0.042)	0.001 (0.041)	0.000 (0.030)
Standard IPW	0.001 (0.077)	$-$ 0.006 (0.069)	0.002 (0.050)
Scenario 2: No direct effect of treatment, delivery time has effect
True value	0.028	0.027	0
IPW	0.003 (0.051)	$-$ 0.003 (0.046)	$-$ 0.001 (0.029)
G-computation	$-$ 0.003 (0.051)	0.000 (0.043)	0.000 (0.029)
TMLE	0.001 (0.051)	$-$ 0.001 (0.045)	0.000 (0.030)
Standard IPW	$-$ 0.027 (0.079)	$-$ 0.031 (0.069)	0.000 (0.051)
Scenario 3: Direct effect of early treatment ( $A(2)$ ), delivery time also has effect
True value	0.110	0.026	0
IPW	$-$ 0.006 (0.054)	0.002 (0.045)	$-$ 0.002 (0.030)
G-computation	$-$ 0.002 (0.051)	$-$ 0.002 (0.043)	0.000 (0.029)
TMLE	$-$ 0.007 (0.053)	0.002 (0.046)	$-$ 0.002 (0.030)
Standard IPW	$-$ 0.061 (0.080)	$-$ 0.029 (0.069)	$-$ 0.001 (0.050)
Scenario 4: Direct effect of late treatment ( $A(4)$ ), delivery time also has effect
True value	0.025	0.012	0.032
TS-IPW	$-$ 0.000 (0.051)	0.001 (0.048)	0.000 (0.035)
TS-G-computation	0.001 (0.050)	0.000 (0.047)	0.001 (0.031)
TS-TMLE	0.002 (0.051)	0.000 (0.048)	0.000 (0.034)
Standard IPW	$-$ 0.027 (0.090)	$-$ 0.016 (0.081)	0.021 (0.056)