. Author manuscript; available in PMC: 2016 Dec 10.

Published in final edited form as: Stat Med. 2015 Aug 2;34(28):3637–3647. doi: 10.1002/sim.6618

Table 2.

Comparison of absolute bias of the estimates of mean Y between various weighting adjustments under four response propensity models in the simulation study: Y₁|X, Z ~ Norm(0, 1), Y₂|X, Z ~ Norm(X, 1), and Y₃|X, Z ~ Norm(X + Z, 1).

Response model	R₁			R₂			R₃			R₄
Outcome	Y₁	Y₂	Y₃	Y₁	Y₂	Y₃	Y₁	Y₂	Y₃	Y₁	Y₂	Y₃
NULL	.000	.000	.005	.024	.363	.344	.016	.313	.407	.014	.139	.152
CHAID_[_x_]	.000	.001	.005	.014	.033	.040	.006	.161	.245	.011	.055	.024
CHAID_[_x,z_]	.000	.001	.006	.014	.033	.041	.016	.076	.078	.000	.003	.003
RP_[_x_]	.000	.000	.005	.001	.003	.002	.001	.152	.232	.002	.055	.028
RP_[_x,z_]	.000	.001	.002	.001	.006	.000	.001	.012	.004	.016	.042	.035
RPS_[_x_]	.000	.000	.004	.001	.029	.022	.003	.165	.234	.004	.038	.018
RPS_[_x,z_]	.000	.001	.003	.004	.035	.020	.000	.078	.040	.016	.039	.033
Hybrid_[_x,z_]	.000	.000	.004	.002	.018	.015	.001	.069	.064	.003	.008	.014