Skip to main content
. 2023 Oct 11;193(2):377–388. doi: 10.1093/aje/kwad196

Table 1.

Performance of Different Estimators of the ATT Based on the Disease Risk Score, Derived Using Simulated Data (1,000 Replicated Samples)

Confounding Status Effect Modification Estimator Bias (×100) MSE (×100)
Weak No Naive −0.5617 0.1176
Weak No NNM 0.0046 0.1575
Weak No OFM 0.1013 0.1419
Weak No IPW −0.0659 0.1078
Weak No TDW −0.1076 0.1086
Moderate No Naive −1.7546 0.1435
Moderate No NNM −0.1308 0.1251
Moderate No OFM 0.0130 0.1102
Moderate No IPW 0.0313 0.0862
Moderate No TDW −0.0968 0.0873
Strong No Naive −3.0799 0.1994
Strong No NNM −0.1081 0.0910
Strong No OFM −0.0664 0.0825
Strong No IPW −0.0665 0.0688
Strong No TDW 0.0687 0.0702
Weak Yes Naive −0.5617 0.1250
Weak Yes NNM 0.0301 0.1671
Weak Yes OFM 0.1013 0.1557
Weak Yes IPW −0.0659 0.1159
Weak Yes TDW −0.1076 0.1165
Moderate Yes Naive −1.7546 0.1487
Moderate Yes NNM −0.1001 0.1344
Moderate Yes OFM 0.0130 0.1175
Moderate Yes IPW 0.0313 0.0953
Moderate Yes TDW −0.0968 0.0965
Strong Yes Naive −3.0799 0.1986
Strong Yes NNM −0.1496 0.0893
Strong Yes OFM −0.0664 0.0815
Strong Yes IPW −0.0665 0.0701
Strong Yes TDW 0.0687 0.0710

Abbreviations: ATT, average treatment effect in the treated population; IPW, inverse probability weighting; MSE, mean squared error; NNM, nearest-neighbor matching; OFM, optimal full matching; TDW, target distribution weighting.