Table 1.

Simulation Experiment 1: results are based on an analysis in which the PS is controlled for in a multivariate model using a parametric spline. We report the estimated bias, variance, and mean-squared error (MSE) of all possible estimators and the average C statistic of the corresponding PS model.

Variables in Propensity Score Model
		X1	X2	X3	X1, X2	X1, X3	X2, X3	X1, X2, X3	None
n=500	$B\widehat{I}AS\times {10}^1$	−0.03	5.97	7.34	−0.03	−0.07	7.36	−0.06	5.94
	$V\widehat{A}R\times {10}^1$	0.32	0.22	0.46	0.22	0.44	0.36	0.31	0.39
	$M\widehat{S}E\times {10}^1$	0.32	3.79	5.85	0.22	0.44	5.77	0.31	3.92
	Avg. C-stat.	0.67	0.52	0.76	0.67	0.82	0.76	0.82
n=2500	$B\widehat{I}AS\times {10}^1$	0.00	5.93	7.33	−0.01	−0.04	7.33	−0.03	5.95
	$V\widehat{A}R\times {10}^2$	0.66	0.53	0.96	0.49	0.89	0.79	0.69	0.80
	$M\widehat{S}E\times {10}^2$	0.66	35.65	54.72	0.49	0.89	54.56	0.70	36.16
	Avg. C-stat.	0.67	0.51	0.76	0.67	0.81	0.76	0.81