Table 1.
Distribution of Propensity Score Weights for the Unexposed Group by Estimation Method and True Propensity Score Model Scenario.
Table 2.
Spearman Correlations of Estimated Propensity Score Weights by Estimation Method and True Propensity Score Model Scenario.
Figure 1.
Average percent absolute bias in the estimate of treatment effect after propensity score weight trimming for 1000 simulated datasets of N = 500, by propensity score estimation method and degree of complexity in the true propensity score model scenario.
Scenario 1: additivity and linearity; Scenario 2: mild non-additivity and non-linearity; Scenario 3: moderate non-additivity and non-linearity. The 100th percentile of weight trimming indicates no trimming was applied.
Figure 2.
Average standard error in the estimate of treatment effect after propensity score weight trimming for 1000 simulated datasets of N = 500, by propensity score estimation method and degree of complexity in the true propensity score model scenario.
Scenario 1: additivity and linearity; Scenario 2: mild non-additivity and non-linearity; Scenario 3: moderate non-additivity and non-linearity. The 100th percentile of weight trimming indicates no trimming was applied.
Figure 3.
95% confidence interval coverage for 1000 simulated datasets of N = 500 after propensity score weight trimming, by propensity score estimation method and degree of complexity in the true propensity score model scenario.
Scenario 1: additivity and linearity; Scenario 2: mild non-additivity and non-linearity; Scenario 3: moderate non-additivity and non-linearity. The 100th percentile of weight trimming indicates no trimming was applied.