Table 3.

Calibration Simulations: Standardized bias

Scenario	1	2	3	4	5	6	7	8	avg bias	avg bias
	Ignorable				Non-ignorable				Ignorable	Non-ignorable
OLS (T)	0.00	0.09	−0.01	−0.06	−0.05	0.16	0.13	0.58	0.00	0.20
BART (T)	−0.02	0.04	−0.02	−0.01	0.06	0.47	0.17	0.84	0.00	0.39
OLS	0.00	0.13	0.00	−0.02	−0.13	0.08	0.05	0.49	0.03	0.12
BART	−0.04	0.06	0.00	−0.02	0.05	0.31	0.01	0.65	0.00	0.25
IPSW-LR	−0.09	−0.11	0.01	0.04	−0.17	0.49	−0.01	0.76	−0.04	0.27
IPSW-RF	−0.09	−0.03	0.03	0.00	−0.10	0.43	0.07	0.69	−0.02	0.27
IPSW-GBM	−0.10	−0.11	0.03	0.05	−0.14	0.61	0.06	0.79	−0.03	0.33
DR-LR	0.01	0.12	−0.02	0.01	−0.21	0.16	0.05	0.54	0.03	0.14
DR-RF	0.00	0.17	−0.01	0.00	−0.18	0.10	0.07	0.50	0.04	0.12
DR-GBM	0.02	0.14	0.01	0.06	−0.15	0.21	0.08	0.55	0.06	0.17

Note: All results reported here average over 10,000 simulated datasets. See Figures 1 and 2 for a description of the scenarios. Standardized Bias refers to bias divided by the standard deviation of the outcome in the target population. (T) refers to simulations in which control outcomes are available in the target dataset. OLS denotes linear regression; BART denotes Bayesian Additive Regression Trees; IPSW-LR (IPSW-RF/IPSW-GBM) denotes inverse propensity score weighting with propensity scores estimated using logistic regression (random forests/boosting); DR-LR (DR-RF/DR-GBM) refers to double robust weighted linear regression models with propensity scores estimated using logistic regression (random forests/boosting). The last two columns show the average standardized bias for the ignorable (1–4) and non-ignorable (5–8) scenarios.