TABLE 3.
Treatment Effect (Trt) and Standard Error (SE) Estimates in the Case Study (Trt(SE))
Method | Stratification | Opt. Matching | Regression | Weighting |
---|---|---|---|---|
Propensity score methods | ||||
BMA-Approx+Bayes | ||||
Occam’s window on | 2.23(.75) | 2.56(.86) | 2.51(.78) | NA |
Occam’s window off | 2.23(.75) | 2.52(.83) | 2.51(.78) | NA |
BMA-Approx+OLS | ||||
Occam’s window on | 2.25(.78) | 2.55(.83) | 2.51(.79) | 2.52(.76) |
Occam’s window off | 2.25(.78) | 2.48(.86) | 2.51(.79) | 2.52(.76) |
BMA-MCMC+Bayes | ||||
Occam’s window on | 2.08(.81) | 2.42(.87) | 2.47(.81) | NA |
BMA-MCMC+OLS | ||||
Occam’s window on | 2.09(.84) | 2.42(.86) | 2.47(.82) | 2.57(.84) |
Two-step BPSA | 2.02(.83) | 2.22(.95) | 2.23(.84) | NA |
Other methods | Estimate(S.E.) | |||
Bayes unadjusted regression | 1.74(.77) | |||
Bayes covariate-adjusted regression | 2.30(.72) |
Note. BPSA = Bayesian propensity score analysis; BMA-Approx+OLS = Approximate Bayesian model averaging with ordinary least squares outcome model; BMA-Approx+Bayes = Approximate Bayesian model averaging with Bayesian outcome model; BMA-MCMC+OLS = Fully Bayesian model averaging with OLS outcome model; BMA-MCMC+Bayes = Fully Bayesian model averaging with Bayesian outcome model.