Table 1 .

Performance of Rubin’s Rules in simulation study

		Mean	Mean		Sqrt mean		RR cover
“No treatment effect” scenario
MAR	0.0	−0.009	−0.009	0.784	0.820	0.823	0.948
copy ref	0.0	−0.009	−0.006	0.783	0.818	0.700	0.977
jump to ref	0.0	−0.009	−0.007	0.784	0.827	0.663	0.984
copy increm	0.0	−0.009	−0.007	0.784	0.823	0.715	0.974
LMCF	1.6	1.592	1.594	0.846	0.876	0.828	0.961
“Treatment effect” scenario
MAR	−3.0	−3.019	−3.020	0.778	0.820	0.818	0.948
copy ref	−2.4	−2.417	−2.415	0.786	0.827	0.708	0.975
jump to ref	−2.4	−2.417	−2.415	0.787	0.835	0.668	0.983
copy increm	−2.8	−2.818	−2.815	0.779	0.823	0.715	0.975
LMCF	−1.2	−1.214	−1.213	0.856	0.892	0.842	0.959

Note. is true treatment effect; mean is average of complete-data estimates of (maximum Monte Carlo standard error = 0.0086); mean is average of RR treatment effect estimates (max MCSE = 0.0084); is empirical standard error of complete-data estimates (max MCSE = 0.0061); sqrt mean is square root of the average RR estimate of the variance (max MCSE = 0.0005); is empirical standard error of RR estimate (max MCSE = 0.0060); RR cover is coverage of 95% confidence interval from Rubin’s Rules (max MCSE = 0.0022).